aboutsummaryrefslogtreecommitdiff
path: root/fs/dlm
diff options
context:
space:
mode:
Diffstat (limited to 'fs/dlm')
-rw-r--r--fs/dlm/Kconfig5
-rw-r--r--fs/dlm/ast.c393
-rw-r--r--fs/dlm/ast.h18
-rw-r--r--fs/dlm/config.c317
-rw-r--r--fs/dlm/config.h23
-rw-r--r--fs/dlm/debug_fs.c162
-rw-r--r--fs/dlm/dir.c297
-rw-r--r--fs/dlm/dir.h7
-rw-r--r--fs/dlm/dlm_internal.h251
-rw-r--r--fs/dlm/lock.c2263
-rw-r--r--fs/dlm/lock.h14
-rw-r--r--fs/dlm/lockspace.c355
-rw-r--r--fs/dlm/lowcomms.c533
-rw-r--r--fs/dlm/lowcomms.h2
-rw-r--r--fs/dlm/main.c4
-rw-r--r--fs/dlm/member.c506
-rw-r--r--fs/dlm/member.h10
-rw-r--r--fs/dlm/memory.c28
-rw-r--r--fs/dlm/memory.h2
-rw-r--r--fs/dlm/netlink.c18
-rw-r--r--fs/dlm/plock.c91
-rw-r--r--fs/dlm/rcom.c271
-rw-r--r--fs/dlm/rcom.h3
-rw-r--r--fs/dlm/recover.c448
-rw-r--r--fs/dlm/recover.h2
-rw-r--r--fs/dlm/recoverd.c115
-rw-r--r--fs/dlm/recoverd.h1
-rw-r--r--fs/dlm/requestqueue.c43
-rw-r--r--fs/dlm/user.c229
-rw-r--r--fs/dlm/user.h3
30 files changed, 4449 insertions, 1965 deletions
diff --git a/fs/dlm/Kconfig b/fs/dlm/Kconfig
index 2dbb422e811..e4242c3f848 100644
--- a/fs/dlm/Kconfig
+++ b/fs/dlm/Kconfig
@@ -1,8 +1,7 @@
menuconfig DLM
tristate "Distributed Lock Manager (DLM)"
- depends on EXPERIMENTAL && INET
- depends on SYSFS && (IPV6 || IPV6=n)
- select CONFIGFS_FS
+ depends on INET
+ depends on SYSFS && CONFIGFS_FS && (IPV6 || IPV6=n)
select IP_SCTP
help
A general purpose distributed lock manager for kernel or userspace
diff --git a/fs/dlm/ast.c b/fs/dlm/ast.c
index 4314f0d48d8..dcea1e37a1b 100644
--- a/fs/dlm/ast.c
+++ b/fs/dlm/ast.c
@@ -16,192 +16,299 @@
#include "user.h"
#include "ast.h"
-#define WAKE_ASTS 0
+static uint64_t dlm_cb_seq;
+static DEFINE_SPINLOCK(dlm_cb_seq_spin);
-static struct list_head ast_queue;
-static spinlock_t ast_queue_lock;
-static struct task_struct * astd_task;
-static unsigned long astd_wakeflags;
-static struct mutex astd_running;
-
-
-void dlm_del_ast(struct dlm_lkb *lkb)
+static void dlm_dump_lkb_callbacks(struct dlm_lkb *lkb)
{
- spin_lock(&ast_queue_lock);
- if (lkb->lkb_ast_type & (AST_COMP | AST_BAST))
- list_del(&lkb->lkb_astqueue);
- spin_unlock(&ast_queue_lock);
+ int i;
+
+ log_print("last_bast %x %llu flags %x mode %d sb %d %x",
+ lkb->lkb_id,
+ (unsigned long long)lkb->lkb_last_bast.seq,
+ lkb->lkb_last_bast.flags,
+ lkb->lkb_last_bast.mode,
+ lkb->lkb_last_bast.sb_status,
+ lkb->lkb_last_bast.sb_flags);
+
+ log_print("last_cast %x %llu flags %x mode %d sb %d %x",
+ lkb->lkb_id,
+ (unsigned long long)lkb->lkb_last_cast.seq,
+ lkb->lkb_last_cast.flags,
+ lkb->lkb_last_cast.mode,
+ lkb->lkb_last_cast.sb_status,
+ lkb->lkb_last_cast.sb_flags);
+
+ for (i = 0; i < DLM_CALLBACKS_SIZE; i++) {
+ log_print("cb %x %llu flags %x mode %d sb %d %x",
+ lkb->lkb_id,
+ (unsigned long long)lkb->lkb_callbacks[i].seq,
+ lkb->lkb_callbacks[i].flags,
+ lkb->lkb_callbacks[i].mode,
+ lkb->lkb_callbacks[i].sb_status,
+ lkb->lkb_callbacks[i].sb_flags);
+ }
}
-void dlm_add_ast(struct dlm_lkb *lkb, int type, int mode)
+int dlm_add_lkb_callback(struct dlm_lkb *lkb, uint32_t flags, int mode,
+ int status, uint32_t sbflags, uint64_t seq)
{
- if (lkb->lkb_flags & DLM_IFL_USER) {
- dlm_user_add_ast(lkb, type, mode);
- return;
+ struct dlm_ls *ls = lkb->lkb_resource->res_ls;
+ uint64_t prev_seq;
+ int prev_mode;
+ int i, rv;
+
+ for (i = 0; i < DLM_CALLBACKS_SIZE; i++) {
+ if (lkb->lkb_callbacks[i].seq)
+ continue;
+
+ /*
+ * Suppress some redundant basts here, do more on removal.
+ * Don't even add a bast if the callback just before it
+ * is a bast for the same mode or a more restrictive mode.
+ * (the addional > PR check is needed for PR/CW inversion)
+ */
+
+ if ((i > 0) && (flags & DLM_CB_BAST) &&
+ (lkb->lkb_callbacks[i-1].flags & DLM_CB_BAST)) {
+
+ prev_seq = lkb->lkb_callbacks[i-1].seq;
+ prev_mode = lkb->lkb_callbacks[i-1].mode;
+
+ if ((prev_mode == mode) ||
+ (prev_mode > mode && prev_mode > DLM_LOCK_PR)) {
+
+ log_debug(ls, "skip %x add bast %llu mode %d "
+ "for bast %llu mode %d",
+ lkb->lkb_id,
+ (unsigned long long)seq,
+ mode,
+ (unsigned long long)prev_seq,
+ prev_mode);
+ rv = 0;
+ goto out;
+ }
+ }
+
+ lkb->lkb_callbacks[i].seq = seq;
+ lkb->lkb_callbacks[i].flags = flags;
+ lkb->lkb_callbacks[i].mode = mode;
+ lkb->lkb_callbacks[i].sb_status = status;
+ lkb->lkb_callbacks[i].sb_flags = (sbflags & 0x000000FF);
+ rv = 0;
+ break;
}
- spin_lock(&ast_queue_lock);
- if (!(lkb->lkb_ast_type & (AST_COMP | AST_BAST))) {
- kref_get(&lkb->lkb_ref);
- list_add_tail(&lkb->lkb_astqueue, &ast_queue);
- lkb->lkb_ast_first = type;
+ if (i == DLM_CALLBACKS_SIZE) {
+ log_error(ls, "no callbacks %x %llu flags %x mode %d sb %d %x",
+ lkb->lkb_id, (unsigned long long)seq,
+ flags, mode, status, sbflags);
+ dlm_dump_lkb_callbacks(lkb);
+ rv = -1;
+ goto out;
+ }
+ out:
+ return rv;
+}
+
+int dlm_rem_lkb_callback(struct dlm_ls *ls, struct dlm_lkb *lkb,
+ struct dlm_callback *cb, int *resid)
+{
+ int i, rv;
+
+ *resid = 0;
+
+ if (!lkb->lkb_callbacks[0].seq) {
+ rv = -ENOENT;
+ goto out;
}
- /* sanity check, this should not happen */
+ /* oldest undelivered cb is callbacks[0] */
+
+ memcpy(cb, &lkb->lkb_callbacks[0], sizeof(struct dlm_callback));
+ memset(&lkb->lkb_callbacks[0], 0, sizeof(struct dlm_callback));
+
+ /* shift others down */
- if ((type == AST_COMP) && (lkb->lkb_ast_type & AST_COMP))
- log_print("repeat cast %d castmode %d lock %x %s",
- mode, lkb->lkb_castmode,
- lkb->lkb_id, lkb->lkb_resource->res_name);
+ for (i = 1; i < DLM_CALLBACKS_SIZE; i++) {
+ if (!lkb->lkb_callbacks[i].seq)
+ break;
+ memcpy(&lkb->lkb_callbacks[i-1], &lkb->lkb_callbacks[i],
+ sizeof(struct dlm_callback));
+ memset(&lkb->lkb_callbacks[i], 0, sizeof(struct dlm_callback));
+ (*resid)++;
+ }
- lkb->lkb_ast_type |= type;
- if (type == AST_BAST)
- lkb->lkb_bastmode = mode;
- else
- lkb->lkb_castmode = mode;
- spin_unlock(&ast_queue_lock);
+ /* if cb is a bast, it should be skipped if the blocking mode is
+ compatible with the last granted mode */
+
+ if ((cb->flags & DLM_CB_BAST) && lkb->lkb_last_cast.seq) {
+ if (dlm_modes_compat(cb->mode, lkb->lkb_last_cast.mode)) {
+ cb->flags |= DLM_CB_SKIP;
+
+ log_debug(ls, "skip %x bast %llu mode %d "
+ "for cast %llu mode %d",
+ lkb->lkb_id,
+ (unsigned long long)cb->seq,
+ cb->mode,
+ (unsigned long long)lkb->lkb_last_cast.seq,
+ lkb->lkb_last_cast.mode);
+ rv = 0;
+ goto out;
+ }
+ }
- set_bit(WAKE_ASTS, &astd_wakeflags);
- wake_up_process(astd_task);
+ if (cb->flags & DLM_CB_CAST) {
+ memcpy(&lkb->lkb_last_cast, cb, sizeof(struct dlm_callback));
+ lkb->lkb_last_cast_time = ktime_get();
+ }
+
+ if (cb->flags & DLM_CB_BAST) {
+ memcpy(&lkb->lkb_last_bast, cb, sizeof(struct dlm_callback));
+ lkb->lkb_last_bast_time = ktime_get();
+ }
+ rv = 0;
+ out:
+ return rv;
}
-static void process_asts(void)
+void dlm_add_cb(struct dlm_lkb *lkb, uint32_t flags, int mode, int status,
+ uint32_t sbflags)
{
- struct dlm_ls *ls = NULL;
- struct dlm_rsb *r = NULL;
- struct dlm_lkb *lkb;
- void (*castfn) (void *astparam);
- void (*bastfn) (void *astparam, int mode);
- int type, first, bastmode, castmode, do_bast, do_cast, last_castmode;
+ struct dlm_ls *ls = lkb->lkb_resource->res_ls;
+ uint64_t new_seq, prev_seq;
+ int rv;
-repeat:
- spin_lock(&ast_queue_lock);
- list_for_each_entry(lkb, &ast_queue, lkb_astqueue) {
- r = lkb->lkb_resource;
- ls = r->res_ls;
+ spin_lock(&dlm_cb_seq_spin);
+ new_seq = ++dlm_cb_seq;
+ spin_unlock(&dlm_cb_seq_spin);
- if (dlm_locking_stopped(ls))
- continue;
+ if (lkb->lkb_flags & DLM_IFL_USER) {
+ dlm_user_add_ast(lkb, flags, mode, status, sbflags, new_seq);
+ return;
+ }
- list_del(&lkb->lkb_astqueue);
- type = lkb->lkb_ast_type;
- lkb->lkb_ast_type = 0;
- first = lkb->lkb_ast_first;
- lkb->lkb_ast_first = 0;
- bastmode = lkb->lkb_bastmode;
- castmode = lkb->lkb_castmode;
- castfn = lkb->lkb_astfn;
- bastfn = lkb->lkb_bastfn;
- spin_unlock(&ast_queue_lock);
-
- do_cast = (type & AST_COMP) && castfn;
- do_bast = (type & AST_BAST) && bastfn;
-
- /* Skip a bast if its blocking mode is compatible with the
- granted mode of the preceding cast. */
-
- if (do_bast) {
- if (first == AST_COMP)
- last_castmode = castmode;
- else
- last_castmode = lkb->lkb_castmode_done;
- if (dlm_modes_compat(bastmode, last_castmode))
- do_bast = 0;
- }
+ mutex_lock(&lkb->lkb_cb_mutex);
+ prev_seq = lkb->lkb_callbacks[0].seq;
+
+ rv = dlm_add_lkb_callback(lkb, flags, mode, status, sbflags, new_seq);
+ if (rv < 0)
+ goto out;
+
+ if (!prev_seq) {
+ kref_get(&lkb->lkb_ref);
- if (first == AST_COMP) {
- if (do_cast)
- castfn(lkb->lkb_astparam);
- if (do_bast)
- bastfn(lkb->lkb_astparam, bastmode);
- } else if (first == AST_BAST) {
- if (do_bast)
- bastfn(lkb->lkb_astparam, bastmode);
- if (do_cast)
- castfn(lkb->lkb_astparam);
+ if (test_bit(LSFL_CB_DELAY, &ls->ls_flags)) {
+ mutex_lock(&ls->ls_cb_mutex);
+ list_add(&lkb->lkb_cb_list, &ls->ls_cb_delay);
+ mutex_unlock(&ls->ls_cb_mutex);
} else {
- log_error(ls, "bad ast_first %d ast_type %d",
- first, type);
+ queue_work(ls->ls_callback_wq, &lkb->lkb_cb_work);
}
+ }
+ out:
+ mutex_unlock(&lkb->lkb_cb_mutex);
+}
- if (do_cast)
- lkb->lkb_castmode_done = castmode;
- if (do_bast)
- lkb->lkb_bastmode_done = bastmode;
+void dlm_callback_work(struct work_struct *work)
+{
+ struct dlm_lkb *lkb = container_of(work, struct dlm_lkb, lkb_cb_work);
+ struct dlm_ls *ls = lkb->lkb_resource->res_ls;
+ void (*castfn) (void *astparam);
+ void (*bastfn) (void *astparam, int mode);
+ struct dlm_callback callbacks[DLM_CALLBACKS_SIZE];
+ int i, rv, resid;
- /* this removes the reference added by dlm_add_ast
- and may result in the lkb being freed */
- dlm_put_lkb(lkb);
+ memset(&callbacks, 0, sizeof(callbacks));
- cond_resched();
- goto repeat;
+ mutex_lock(&lkb->lkb_cb_mutex);
+ if (!lkb->lkb_callbacks[0].seq) {
+ /* no callback work exists, shouldn't happen */
+ log_error(ls, "dlm_callback_work %x no work", lkb->lkb_id);
+ dlm_print_lkb(lkb);
+ dlm_dump_lkb_callbacks(lkb);
}
- spin_unlock(&ast_queue_lock);
-}
-static inline int no_asts(void)
-{
- int ret;
+ for (i = 0; i < DLM_CALLBACKS_SIZE; i++) {
+ rv = dlm_rem_lkb_callback(ls, lkb, &callbacks[i], &resid);
+ if (rv < 0)
+ break;
+ }
- spin_lock(&ast_queue_lock);
- ret = list_empty(&ast_queue);
- spin_unlock(&ast_queue_lock);
- return ret;
-}
+ if (resid) {
+ /* cbs remain, loop should have removed all, shouldn't happen */
+ log_error(ls, "dlm_callback_work %x resid %d", lkb->lkb_id,
+ resid);
+ dlm_print_lkb(lkb);
+ dlm_dump_lkb_callbacks(lkb);
+ }
+ mutex_unlock(&lkb->lkb_cb_mutex);
-static int dlm_astd(void *data)
-{
- while (!kthread_should_stop()) {
- set_current_state(TASK_INTERRUPTIBLE);
- if (!test_bit(WAKE_ASTS, &astd_wakeflags))
- schedule();
- set_current_state(TASK_RUNNING);
+ castfn = lkb->lkb_astfn;
+ bastfn = lkb->lkb_bastfn;
- mutex_lock(&astd_running);
- if (test_and_clear_bit(WAKE_ASTS, &astd_wakeflags))
- process_asts();
- mutex_unlock(&astd_running);
+ for (i = 0; i < DLM_CALLBACKS_SIZE; i++) {
+ if (!callbacks[i].seq)
+ break;
+ if (callbacks[i].flags & DLM_CB_SKIP) {
+ continue;
+ } else if (callbacks[i].flags & DLM_CB_BAST) {
+ bastfn(lkb->lkb_astparam, callbacks[i].mode);
+ } else if (callbacks[i].flags & DLM_CB_CAST) {
+ lkb->lkb_lksb->sb_status = callbacks[i].sb_status;
+ lkb->lkb_lksb->sb_flags = callbacks[i].sb_flags;
+ castfn(lkb->lkb_astparam);
+ }
}
- return 0;
+
+ /* undo kref_get from dlm_add_callback, may cause lkb to be freed */
+ dlm_put_lkb(lkb);
}
-void dlm_astd_wake(void)
+int dlm_callback_start(struct dlm_ls *ls)
{
- if (!no_asts()) {
- set_bit(WAKE_ASTS, &astd_wakeflags);
- wake_up_process(astd_task);
+ ls->ls_callback_wq = alloc_workqueue("dlm_callback",
+ WQ_UNBOUND | WQ_MEM_RECLAIM, 0);
+ if (!ls->ls_callback_wq) {
+ log_print("can't start dlm_callback workqueue");
+ return -ENOMEM;
}
+ return 0;
}
-int dlm_astd_start(void)
+void dlm_callback_stop(struct dlm_ls *ls)
{
- struct task_struct *p;
- int error = 0;
-
- INIT_LIST_HEAD(&ast_queue);
- spin_lock_init(&ast_queue_lock);
- mutex_init(&astd_running);
-
- p = kthread_run(dlm_astd, NULL, "dlm_astd");
- if (IS_ERR(p))
- error = PTR_ERR(p);
- else
- astd_task = p;
- return error;
+ if (ls->ls_callback_wq)
+ destroy_workqueue(ls->ls_callback_wq);
}
-void dlm_astd_stop(void)
+void dlm_callback_suspend(struct dlm_ls *ls)
{
- kthread_stop(astd_task);
-}
+ set_bit(LSFL_CB_DELAY, &ls->ls_flags);
-void dlm_astd_suspend(void)
-{
- mutex_lock(&astd_running);
+ if (ls->ls_callback_wq)
+ flush_workqueue(ls->ls_callback_wq);
}
-void dlm_astd_resume(void)
+void dlm_callback_resume(struct dlm_ls *ls)
{
- mutex_unlock(&astd_running);
+ struct dlm_lkb *lkb, *safe;
+ int count = 0;
+
+ clear_bit(LSFL_CB_DELAY, &ls->ls_flags);
+
+ if (!ls->ls_callback_wq)
+ return;
+
+ mutex_lock(&ls->ls_cb_mutex);
+ list_for_each_entry_safe(lkb, safe, &ls->ls_cb_delay, lkb_cb_list) {
+ list_del_init(&lkb->lkb_cb_list);
+ queue_work(ls->ls_callback_wq, &lkb->lkb_cb_work);
+ count++;
+ }
+ mutex_unlock(&ls->ls_cb_mutex);
+
+ if (count)
+ log_rinfo(ls, "dlm_callback_resume %d", count);
}
diff --git a/fs/dlm/ast.h b/fs/dlm/ast.h
index bcb1aaba519..757b551c682 100644
--- a/fs/dlm/ast.h
+++ b/fs/dlm/ast.h
@@ -13,14 +13,20 @@
#ifndef __ASTD_DOT_H__
#define __ASTD_DOT_H__
-void dlm_add_ast(struct dlm_lkb *lkb, int type, int mode);
void dlm_del_ast(struct dlm_lkb *lkb);
+int dlm_add_lkb_callback(struct dlm_lkb *lkb, uint32_t flags, int mode,
+ int status, uint32_t sbflags, uint64_t seq);
+int dlm_rem_lkb_callback(struct dlm_ls *ls, struct dlm_lkb *lkb,
+ struct dlm_callback *cb, int *resid);
+void dlm_add_cb(struct dlm_lkb *lkb, uint32_t flags, int mode, int status,
+ uint32_t sbflags);
-void dlm_astd_wake(void);
-int dlm_astd_start(void);
-void dlm_astd_stop(void);
-void dlm_astd_suspend(void);
-void dlm_astd_resume(void);
+void dlm_callback_work(struct work_struct *work);
+int dlm_callback_start(struct dlm_ls *ls);
+void dlm_callback_stop(struct dlm_ls *ls);
+void dlm_callback_suspend(struct dlm_ls *ls);
+void dlm_callback_resume(struct dlm_ls *ls);
#endif
+
diff --git a/fs/dlm/config.c b/fs/dlm/config.c
index b54bca03d92..d521bddf876 100644
--- a/fs/dlm/config.c
+++ b/fs/dlm/config.c
@@ -2,7 +2,7 @@
*******************************************************************************
**
** Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
-** Copyright (C) 2004-2008 Red Hat, Inc. All rights reserved.
+** Copyright (C) 2004-2011 Red Hat, Inc. All rights reserved.
**
** This copyrighted material is made available to anyone wishing to use,
** modify, copy, or redistribute it subject to the terms and conditions
@@ -17,6 +17,7 @@
#include <linux/slab.h>
#include <linux/in.h>
#include <linux/in6.h>
+#include <linux/dlmconstants.h>
#include <net/ipv6.h>
#include <net/sock.h>
@@ -28,13 +29,15 @@
* /config/dlm/<cluster>/spaces/<space>/nodes/<node>/weight
* /config/dlm/<cluster>/comms/<comm>/nodeid
* /config/dlm/<cluster>/comms/<comm>/local
- * /config/dlm/<cluster>/comms/<comm>/addr
+ * /config/dlm/<cluster>/comms/<comm>/addr (write only)
+ * /config/dlm/<cluster>/comms/<comm>/addr_list (read only)
* The <cluster> level is useless, but I haven't figured out how to avoid it.
*/
static struct config_group *space_list;
static struct config_group *comm_list;
static struct dlm_comm *local_comm;
+static uint32_t dlm_comm_count;
struct dlm_clusters;
struct dlm_cluster;
@@ -80,6 +83,7 @@ static ssize_t comm_local_write(struct dlm_comm *cm, const char *buf,
size_t len);
static ssize_t comm_addr_write(struct dlm_comm *cm, const char *buf,
size_t len);
+static ssize_t comm_addr_list_read(struct dlm_comm *cm, char *buf);
static ssize_t node_nodeid_read(struct dlm_node *nd, char *buf);
static ssize_t node_nodeid_write(struct dlm_node *nd, const char *buf,
size_t len);
@@ -92,28 +96,32 @@ struct dlm_cluster {
unsigned int cl_tcp_port;
unsigned int cl_buffer_size;
unsigned int cl_rsbtbl_size;
- unsigned int cl_lkbtbl_size;
- unsigned int cl_dirtbl_size;
unsigned int cl_recover_timer;
unsigned int cl_toss_secs;
unsigned int cl_scan_secs;
unsigned int cl_log_debug;
unsigned int cl_protocol;
unsigned int cl_timewarn_cs;
+ unsigned int cl_waitwarn_us;
+ unsigned int cl_new_rsb_count;
+ unsigned int cl_recover_callbacks;
+ char cl_cluster_name[DLM_LOCKSPACE_LEN];
};
enum {
CLUSTER_ATTR_TCP_PORT = 0,
CLUSTER_ATTR_BUFFER_SIZE,
CLUSTER_ATTR_RSBTBL_SIZE,
- CLUSTER_ATTR_LKBTBL_SIZE,
- CLUSTER_ATTR_DIRTBL_SIZE,
CLUSTER_ATTR_RECOVER_TIMER,
CLUSTER_ATTR_TOSS_SECS,
CLUSTER_ATTR_SCAN_SECS,
CLUSTER_ATTR_LOG_DEBUG,
CLUSTER_ATTR_PROTOCOL,
CLUSTER_ATTR_TIMEWARN_CS,
+ CLUSTER_ATTR_WAITWARN_US,
+ CLUSTER_ATTR_NEW_RSB_COUNT,
+ CLUSTER_ATTR_RECOVER_CALLBACKS,
+ CLUSTER_ATTR_CLUSTER_NAME,
};
struct cluster_attribute {
@@ -122,16 +130,40 @@ struct cluster_attribute {
ssize_t (*store)(struct dlm_cluster *, const char *, size_t);
};
+static ssize_t cluster_cluster_name_read(struct dlm_cluster *cl, char *buf)
+{
+ return sprintf(buf, "%s\n", cl->cl_cluster_name);
+}
+
+static ssize_t cluster_cluster_name_write(struct dlm_cluster *cl,
+ const char *buf, size_t len)
+{
+ strlcpy(dlm_config.ci_cluster_name, buf,
+ sizeof(dlm_config.ci_cluster_name));
+ strlcpy(cl->cl_cluster_name, buf, sizeof(cl->cl_cluster_name));
+ return len;
+}
+
+static struct cluster_attribute cluster_attr_cluster_name = {
+ .attr = { .ca_owner = THIS_MODULE,
+ .ca_name = "cluster_name",
+ .ca_mode = S_IRUGO | S_IWUSR },
+ .show = cluster_cluster_name_read,
+ .store = cluster_cluster_name_write,
+};
+
static ssize_t cluster_set(struct dlm_cluster *cl, unsigned int *cl_field,
int *info_field, int check_zero,
const char *buf, size_t len)
{
unsigned int x;
+ int rc;
if (!capable(CAP_SYS_ADMIN))
- return -EACCES;
-
- x = simple_strtoul(buf, NULL, 0);
+ return -EPERM;
+ rc = kstrtouint(buf, 0, &x);
+ if (rc)
+ return rc;
if (check_zero && !x)
return -EINVAL;
@@ -158,27 +190,30 @@ __CONFIGFS_ATTR(name, 0644, name##_read, name##_write)
CLUSTER_ATTR(tcp_port, 1);
CLUSTER_ATTR(buffer_size, 1);
CLUSTER_ATTR(rsbtbl_size, 1);
-CLUSTER_ATTR(lkbtbl_size, 1);
-CLUSTER_ATTR(dirtbl_size, 1);
CLUSTER_ATTR(recover_timer, 1);
CLUSTER_ATTR(toss_secs, 1);
CLUSTER_ATTR(scan_secs, 1);
CLUSTER_ATTR(log_debug, 0);
CLUSTER_ATTR(protocol, 0);
CLUSTER_ATTR(timewarn_cs, 1);
+CLUSTER_ATTR(waitwarn_us, 0);
+CLUSTER_ATTR(new_rsb_count, 0);
+CLUSTER_ATTR(recover_callbacks, 0);
static struct configfs_attribute *cluster_attrs[] = {
[CLUSTER_ATTR_TCP_PORT] = &cluster_attr_tcp_port.attr,
[CLUSTER_ATTR_BUFFER_SIZE] = &cluster_attr_buffer_size.attr,
[CLUSTER_ATTR_RSBTBL_SIZE] = &cluster_attr_rsbtbl_size.attr,
- [CLUSTER_ATTR_LKBTBL_SIZE] = &cluster_attr_lkbtbl_size.attr,
- [CLUSTER_ATTR_DIRTBL_SIZE] = &cluster_attr_dirtbl_size.attr,
[CLUSTER_ATTR_RECOVER_TIMER] = &cluster_attr_recover_timer.attr,
[CLUSTER_ATTR_TOSS_SECS] = &cluster_attr_toss_secs.attr,
[CLUSTER_ATTR_SCAN_SECS] = &cluster_attr_scan_secs.attr,
[CLUSTER_ATTR_LOG_DEBUG] = &cluster_attr_log_debug.attr,
[CLUSTER_ATTR_PROTOCOL] = &cluster_attr_protocol.attr,
[CLUSTER_ATTR_TIMEWARN_CS] = &cluster_attr_timewarn_cs.attr,
+ [CLUSTER_ATTR_WAITWARN_US] = &cluster_attr_waitwarn_us.attr,
+ [CLUSTER_ATTR_NEW_RSB_COUNT] = &cluster_attr_new_rsb_count.attr,
+ [CLUSTER_ATTR_RECOVER_CALLBACKS] = &cluster_attr_recover_callbacks.attr,
+ [CLUSTER_ATTR_CLUSTER_NAME] = &cluster_attr_cluster_name.attr,
NULL,
};
@@ -186,6 +221,7 @@ enum {
COMM_ATTR_NODEID = 0,
COMM_ATTR_LOCAL,
COMM_ATTR_ADDR,
+ COMM_ATTR_ADDR_LIST,
};
struct comm_attribute {
@@ -213,14 +249,22 @@ static struct comm_attribute comm_attr_local = {
static struct comm_attribute comm_attr_addr = {
.attr = { .ca_owner = THIS_MODULE,
.ca_name = "addr",
- .ca_mode = S_IRUGO | S_IWUSR },
+ .ca_mode = S_IWUSR },
.store = comm_addr_write,
};
+static struct comm_attribute comm_attr_addr_list = {
+ .attr = { .ca_owner = THIS_MODULE,
+ .ca_name = "addr_list",
+ .ca_mode = S_IRUGO },
+ .show = comm_addr_list_read,
+};
+
static struct configfs_attribute *comm_attrs[] = {
[COMM_ATTR_NODEID] = &comm_attr_nodeid.attr,
[COMM_ATTR_LOCAL] = &comm_attr_local.attr,
[COMM_ATTR_ADDR] = &comm_attr_addr.attr,
+ [COMM_ATTR_ADDR_LIST] = &comm_attr_addr_list.attr,
NULL,
};
@@ -278,6 +322,7 @@ struct dlm_comms {
struct dlm_comm {
struct config_item item;
+ int seq;
int nodeid;
int local;
int addr_count;
@@ -294,6 +339,7 @@ struct dlm_node {
int nodeid;
int weight;
int new;
+ int comm_seq; /* copy of cm->seq when nd->nodeid is set */
};
static struct configfs_group_operations clusters_ops = {
@@ -431,14 +477,17 @@ static struct config_group *make_cluster(struct config_group *g,
cl->cl_tcp_port = dlm_config.ci_tcp_port;
cl->cl_buffer_size = dlm_config.ci_buffer_size;
cl->cl_rsbtbl_size = dlm_config.ci_rsbtbl_size;
- cl->cl_lkbtbl_size = dlm_config.ci_lkbtbl_size;
- cl->cl_dirtbl_size = dlm_config.ci_dirtbl_size;
cl->cl_recover_timer = dlm_config.ci_recover_timer;
cl->cl_toss_secs = dlm_config.ci_toss_secs;
cl->cl_scan_secs = dlm_config.ci_scan_secs;
cl->cl_log_debug = dlm_config.ci_log_debug;
cl->cl_protocol = dlm_config.ci_protocol;
cl->cl_timewarn_cs = dlm_config.ci_timewarn_cs;
+ cl->cl_waitwarn_us = dlm_config.ci_waitwarn_us;
+ cl->cl_new_rsb_count = dlm_config.ci_new_rsb_count;
+ cl->cl_recover_callbacks = dlm_config.ci_recover_callbacks;
+ memcpy(cl->cl_cluster_name, dlm_config.ci_cluster_name,
+ DLM_LOCKSPACE_LEN);
space_list = &sps->ss_group;
comm_list = &cms->cs_group;
@@ -542,6 +591,11 @@ static struct config_item *make_comm(struct config_group *g, const char *name)
return ERR_PTR(-ENOMEM);
config_item_init_type_name(&cm->item, name, &comm_type);
+
+ cm->seq = dlm_comm_count++;
+ if (!cm->seq)
+ cm->seq = dlm_comm_count++;
+
cm->nodeid = -1;
cm->local = 0;
cm->addr_count = 0;
@@ -678,7 +732,10 @@ static ssize_t comm_nodeid_read(struct dlm_comm *cm, char *buf)
static ssize_t comm_nodeid_write(struct dlm_comm *cm, const char *buf,
size_t len)
{
- cm->nodeid = simple_strtol(buf, NULL, 0);
+ int rc = kstrtoint(buf, 0, &cm->nodeid);
+
+ if (rc)
+ return rc;
return len;
}
@@ -690,7 +747,10 @@ static ssize_t comm_local_read(struct dlm_comm *cm, char *buf)
static ssize_t comm_local_write(struct dlm_comm *cm, const char *buf,
size_t len)
{
- cm->local= simple_strtol(buf, NULL, 0);
+ int rc = kstrtoint(buf, 0, &cm->local);
+
+ if (rc)
+ return rc;
if (cm->local && !local_comm)
local_comm = cm;
return len;
@@ -699,6 +759,7 @@ static ssize_t comm_local_write(struct dlm_comm *cm, const char *buf,
static ssize_t comm_addr_write(struct dlm_comm *cm, const char *buf, size_t len)
{
struct sockaddr_storage *addr;
+ int rv;
if (len != sizeof(struct sockaddr_storage))
return -EINVAL;
@@ -711,10 +772,61 @@ static ssize_t comm_addr_write(struct dlm_comm *cm, const char *buf, size_t len)
return -ENOMEM;
memcpy(addr, buf, len);
+
+ rv = dlm_lowcomms_addr(cm->nodeid, addr, len);
+ if (rv) {
+ kfree(addr);
+ return rv;
+ }
+
cm->addr[cm->addr_count++] = addr;
return len;
}
+static ssize_t comm_addr_list_read(struct dlm_comm *cm, char *buf)
+{
+ ssize_t s;
+ ssize_t allowance;
+ int i;
+ struct sockaddr_storage *addr;
+ struct sockaddr_in *addr_in;
+ struct sockaddr_in6 *addr_in6;
+
+ /* Taken from ip6_addr_string() defined in lib/vsprintf.c */
+ char buf0[sizeof("AF_INET6 xxxx:xxxx:xxxx:xxxx:xxxx:xxxx:255.255.255.255\n")];
+
+
+ /* Derived from SIMPLE_ATTR_SIZE of fs/configfs/file.c */
+ allowance = 4096;
+ buf[0] = '\0';
+
+ for (i = 0; i < cm->addr_count; i++) {
+ addr = cm->addr[i];
+
+ switch(addr->ss_family) {
+ case AF_INET:
+ addr_in = (struct sockaddr_in *)addr;
+ s = sprintf(buf0, "AF_INET %pI4\n", &addr_in->sin_addr.s_addr);
+ break;
+ case AF_INET6:
+ addr_in6 = (struct sockaddr_in6 *)addr;
+ s = sprintf(buf0, "AF_INET6 %pI6\n", &addr_in6->sin6_addr);
+ break;
+ default:
+ s = sprintf(buf0, "%s\n", "<UNKNOWN>");
+ break;
+ }
+ allowance -= s;
+ if (allowance >= 0)
+ strcat(buf, buf0);
+ else {
+ allowance += s;
+ break;
+ }
+ }
+ return 4096 - allowance;
+}
+
static ssize_t show_node(struct config_item *i, struct configfs_attribute *a,
char *buf)
{
@@ -741,7 +853,13 @@ static ssize_t node_nodeid_read(struct dlm_node *nd, char *buf)
static ssize_t node_nodeid_write(struct dlm_node *nd, const char *buf,
size_t len)
{
- nd->nodeid = simple_strtol(buf, NULL, 0);
+ uint32_t seq = 0;
+ int rc = kstrtoint(buf, 0, &nd->nodeid);
+
+ if (rc)
+ return rc;
+ dlm_comm_seq(nd->nodeid, &seq);
+ nd->comm_seq = seq;
return len;
}
@@ -753,7 +871,10 @@ static ssize_t node_weight_read(struct dlm_node *nd, char *buf)
static ssize_t node_weight_write(struct dlm_node *nd, const char *buf,
size_t len)
{
- nd->weight = simple_strtol(buf, NULL, 0);
+ int rc = kstrtoint(buf, 0, &nd->weight);
+
+ if (rc)
+ return rc;
return len;
}
@@ -780,34 +901,7 @@ static void put_space(struct dlm_space *sp)
config_item_put(&sp->group.cg_item);
}
-static int addr_compare(struct sockaddr_storage *x, struct sockaddr_storage *y)
-{
- switch (x->ss_family) {
- case AF_INET: {
- struct sockaddr_in *sinx = (struct sockaddr_in *)x;
- struct sockaddr_in *siny = (struct sockaddr_in *)y;
- if (sinx->sin_addr.s_addr != siny->sin_addr.s_addr)
- return 0;
- if (sinx->sin_port != siny->sin_port)
- return 0;
- break;
- }
- case AF_INET6: {
- struct sockaddr_in6 *sinx = (struct sockaddr_in6 *)x;
- struct sockaddr_in6 *siny = (struct sockaddr_in6 *)y;
- if (!ipv6_addr_equal(&sinx->sin6_addr, &siny->sin6_addr))
- return 0;
- if (sinx->sin6_port != siny->sin6_port)
- return 0;
- break;
- }
- default:
- return 0;
- }
- return 1;
-}
-
-static struct dlm_comm *get_comm(int nodeid, struct sockaddr_storage *addr)
+static struct dlm_comm *get_comm(int nodeid)
{
struct config_item *i;
struct dlm_comm *cm = NULL;
@@ -821,19 +915,11 @@ static struct dlm_comm *get_comm(int nodeid, struct sockaddr_storage *addr)
list_for_each_entry(i, &comm_list->cg_children, ci_entry) {
cm = config_item_to_comm(i);
- if (nodeid) {
- if (cm->nodeid != nodeid)
- continue;
- found = 1;
- config_item_get(i);
- break;
- } else {
- if (!cm->addr_count || !addr_compare(cm->addr[0], addr))
- continue;
- found = 1;
- config_item_get(i);
- break;
- }
+ if (cm->nodeid != nodeid)
+ continue;
+ found = 1;
+ config_item_get(i);
+ break;
}
mutex_unlock(&clusters_root.subsys.su_mutex);
@@ -848,13 +934,13 @@ static void put_comm(struct dlm_comm *cm)
}
/* caller must free mem */
-int dlm_nodeid_list(char *lsname, int **ids_out, int *ids_count_out,
- int **new_out, int *new_count_out)
+int dlm_config_nodes(char *lsname, struct dlm_config_node **nodes_out,
+ int *count_out)
{
struct dlm_space *sp;
struct dlm_node *nd;
- int i = 0, rv = 0, ids_count = 0, new_count = 0;
- int *ids, *new;
+ struct dlm_config_node *nodes, *node;
+ int rv, count;
sp = get_space(lsname);
if (!sp)
@@ -867,93 +953,40 @@ int dlm_nodeid_list(char *lsname, int **ids_out, int *ids_count_out,
goto out;
}
- ids_count = sp->members_count;
+ count = sp->members_count;
- ids = kcalloc(ids_count, sizeof(int), GFP_NOFS);
- if (!ids) {
+ nodes = kcalloc(count, sizeof(struct dlm_config_node), GFP_NOFS);
+ if (!nodes) {
rv = -ENOMEM;
goto out;
}
+ node = nodes;
list_for_each_entry(nd, &sp->members, list) {
- ids[i++] = nd->nodeid;
- if (nd->new)
- new_count++;
- }
-
- if (ids_count != i)
- printk(KERN_ERR "dlm: bad nodeid count %d %d\n", ids_count, i);
+ node->nodeid = nd->nodeid;
+ node->weight = nd->weight;
+ node->new = nd->new;
+ node->comm_seq = nd->comm_seq;
+ node++;
- if (!new_count)
- goto out_ids;
-
- new = kcalloc(new_count, sizeof(int), GFP_NOFS);
- if (!new) {
- kfree(ids);
- rv = -ENOMEM;
- goto out;
- }
-
- i = 0;
- list_for_each_entry(nd, &sp->members, list) {
- if (nd->new) {
- new[i++] = nd->nodeid;
- nd->new = 0;
- }
+ nd->new = 0;
}
- *new_count_out = new_count;
- *new_out = new;
- out_ids:
- *ids_count_out = ids_count;
- *ids_out = ids;
+ *count_out = count;
+ *nodes_out = nodes;
+ rv = 0;
out:
mutex_unlock(&sp->members_lock);
put_space(sp);
return rv;
}
-int dlm_node_weight(char *lsname, int nodeid)
-{
- struct dlm_space *sp;
- struct dlm_node *nd;
- int w = -EEXIST;
-
- sp = get_space(lsname);
- if (!sp)
- goto out;
-
- mutex_lock(&sp->members_lock);
- list_for_each_entry(nd, &sp->members, list) {
- if (nd->nodeid != nodeid)
- continue;
- w = nd->weight;
- break;
- }
- mutex_unlock(&sp->members_lock);
- put_space(sp);
- out:
- return w;
-}
-
-int dlm_nodeid_to_addr(int nodeid, struct sockaddr_storage *addr)
-{
- struct dlm_comm *cm = get_comm(nodeid, NULL);
- if (!cm)
- return -EEXIST;
- if (!cm->addr_count)
- return -ENOENT;
- memcpy(addr, cm->addr[0], sizeof(*addr));
- put_comm(cm);
- return 0;
-}
-
-int dlm_addr_to_nodeid(struct sockaddr_storage *addr, int *nodeid)
+int dlm_comm_seq(int nodeid, uint32_t *seq)
{
- struct dlm_comm *cm = get_comm(0, addr);
+ struct dlm_comm *cm = get_comm(nodeid);
if (!cm)
return -EEXIST;
- *nodeid = cm->nodeid;
+ *seq = cm->seq;
put_comm(cm);
return 0;
}
@@ -977,27 +1010,31 @@ int dlm_our_addr(struct sockaddr_storage *addr, int num)
/* Config file defaults */
#define DEFAULT_TCP_PORT 21064
#define DEFAULT_BUFFER_SIZE 4096
-#define DEFAULT_RSBTBL_SIZE 256
-#define DEFAULT_LKBTBL_SIZE 1024
-#define DEFAULT_DIRTBL_SIZE 512
+#define DEFAULT_RSBTBL_SIZE 1024
#define DEFAULT_RECOVER_TIMER 5
#define DEFAULT_TOSS_SECS 10
#define DEFAULT_SCAN_SECS 5
#define DEFAULT_LOG_DEBUG 0
#define DEFAULT_PROTOCOL 0
#define DEFAULT_TIMEWARN_CS 500 /* 5 sec = 500 centiseconds */
+#define DEFAULT_WAITWARN_US 0
+#define DEFAULT_NEW_RSB_COUNT 128
+#define DEFAULT_RECOVER_CALLBACKS 0
+#define DEFAULT_CLUSTER_NAME ""
struct dlm_config_info dlm_config = {
.ci_tcp_port = DEFAULT_TCP_PORT,
.ci_buffer_size = DEFAULT_BUFFER_SIZE,
.ci_rsbtbl_size = DEFAULT_RSBTBL_SIZE,
- .ci_lkbtbl_size = DEFAULT_LKBTBL_SIZE,
- .ci_dirtbl_size = DEFAULT_DIRTBL_SIZE,
.ci_recover_timer = DEFAULT_RECOVER_TIMER,
.ci_toss_secs = DEFAULT_TOSS_SECS,
.ci_scan_secs = DEFAULT_SCAN_SECS,
.ci_log_debug = DEFAULT_LOG_DEBUG,
.ci_protocol = DEFAULT_PROTOCOL,
- .ci_timewarn_cs = DEFAULT_TIMEWARN_CS
+ .ci_timewarn_cs = DEFAULT_TIMEWARN_CS,
+ .ci_waitwarn_us = DEFAULT_WAITWARN_US,
+ .ci_new_rsb_count = DEFAULT_NEW_RSB_COUNT,
+ .ci_recover_callbacks = DEFAULT_RECOVER_CALLBACKS,
+ .ci_cluster_name = DEFAULT_CLUSTER_NAME
};
diff --git a/fs/dlm/config.h b/fs/dlm/config.h
index 4f1d6fce58c..f30697bc278 100644
--- a/fs/dlm/config.h
+++ b/fs/dlm/config.h
@@ -2,7 +2,7 @@
*******************************************************************************
**
** Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
-** Copyright (C) 2004-2007 Red Hat, Inc. All rights reserved.
+** Copyright (C) 2004-2011 Red Hat, Inc. All rights reserved.
**
** This copyrighted material is made available to anyone wishing to use,
** modify, copy, or redistribute it subject to the terms and conditions
@@ -14,31 +14,38 @@
#ifndef __CONFIG_DOT_H__
#define __CONFIG_DOT_H__
+struct dlm_config_node {
+ int nodeid;
+ int weight;
+ int new;
+ uint32_t comm_seq;
+};
+
#define DLM_MAX_ADDR_COUNT 3
struct dlm_config_info {
int ci_tcp_port;
int ci_buffer_size;
int ci_rsbtbl_size;
- int ci_lkbtbl_size;
- int ci_dirtbl_size;
int ci_recover_timer;
int ci_toss_secs;
int ci_scan_secs;
int ci_log_debug;
int ci_protocol;
int ci_timewarn_cs;
+ int ci_waitwarn_us;
+ int ci_new_rsb_count;
+ int ci_recover_callbacks;
+ char ci_cluster_name[DLM_LOCKSPACE_LEN];
};
extern struct dlm_config_info dlm_config;
int dlm_config_init(void);
void dlm_config_exit(void);
-int dlm_node_weight(char *lsname, int nodeid);
-int dlm_nodeid_list(char *lsname, int **ids_out, int *ids_count_out,
- int **new_out, int *new_count_out);
-int dlm_nodeid_to_addr(int nodeid, struct sockaddr_storage *addr);
-int dlm_addr_to_nodeid(struct sockaddr_storage *addr, int *nodeid);
+int dlm_config_nodes(char *lsname, struct dlm_config_node **nodes_out,
+ int *count_out);
+int dlm_comm_seq(int nodeid, uint32_t *seq);
int dlm_our_nodeid(void);
int dlm_our_addr(struct sockaddr_storage *addr, int num);
diff --git a/fs/dlm/debug_fs.c b/fs/dlm/debug_fs.c
index 6b42ba807df..8d77ba7b175 100644
--- a/fs/dlm/debug_fs.c
+++ b/fs/dlm/debug_fs.c
@@ -68,7 +68,7 @@ static int print_format1_lock(struct seq_file *s, struct dlm_lkb *lkb,
if (lkb->lkb_wait_type)
seq_printf(s, " wait_type: %d", lkb->lkb_wait_type);
- return seq_printf(s, "\n");
+ return seq_puts(s, "\n");
}
static int print_format1(struct dlm_rsb *res, struct seq_file *s)
@@ -92,31 +92,31 @@ static int print_format1(struct dlm_rsb *res, struct seq_file *s)
}
if (res->res_nodeid > 0)
- rv = seq_printf(s, "\" \nLocal Copy, Master is node %d\n",
+ rv = seq_printf(s, "\"\nLocal Copy, Master is node %d\n",
res->res_nodeid);
else if (res->res_nodeid == 0)
- rv = seq_printf(s, "\" \nMaster Copy\n");
+ rv = seq_puts(s, "\"\nMaster Copy\n");
else if (res->res_nodeid == -1)
- rv = seq_printf(s, "\" \nLooking up master (lkid %x)\n",
+ rv = seq_printf(s, "\"\nLooking up master (lkid %x)\n",
res->res_first_lkid);
else
- rv = seq_printf(s, "\" \nInvalid master %d\n",
+ rv = seq_printf(s, "\"\nInvalid master %d\n",
res->res_nodeid);
if (rv)
goto out;
/* Print the LVB: */
if (res->res_lvbptr) {
- seq_printf(s, "LVB: ");
+ seq_puts(s, "LVB: ");
for (i = 0; i < lvblen; i++) {
if (i == lvblen / 2)
- seq_printf(s, "\n ");
+ seq_puts(s, "\n ");
seq_printf(s, "%02x ",
(unsigned char) res->res_lvbptr[i]);
}
if (rsb_flag(res, RSB_VALNOTVALID))
- seq_printf(s, " (INVALID)");
- rv = seq_printf(s, "\n");
+ seq_puts(s, " (INVALID)");
+ rv = seq_puts(s, "\n");
if (rv)
goto out;
}
@@ -133,21 +133,21 @@ static int print_format1(struct dlm_rsb *res, struct seq_file *s)
}
/* Print the locks attached to this resource */
- seq_printf(s, "Granted Queue\n");
+ seq_puts(s, "Granted Queue\n");
list_for_each_entry(lkb, &res->res_grantqueue, lkb_statequeue) {
rv = print_format1_lock(s, lkb, res);
if (rv)
goto out;
}
- seq_printf(s, "Conversion Queue\n");
+ seq_puts(s, "Conversion Queue\n");
list_for_each_entry(lkb, &res->res_convertqueue, lkb_statequeue) {
rv = print_format1_lock(s, lkb, res);
if (rv)
goto out;
}
- seq_printf(s, "Waiting Queue\n");
+ seq_puts(s, "Waiting Queue\n");
list_for_each_entry(lkb, &res->res_waitqueue, lkb_statequeue) {
rv = print_format1_lock(s, lkb, res);
if (rv)
@@ -157,13 +157,13 @@ static int print_format1(struct dlm_rsb *res, struct seq_file *s)
if (list_empty(&res->res_lookup))
goto out;
- seq_printf(s, "Lookup Queue\n");
+ seq_puts(s, "Lookup Queue\n");
list_for_each_entry(lkb, &res->res_lookup, lkb_rsb_lookup) {
rv = seq_printf(s, "%08x %s", lkb->lkb_id,
print_lockmode(lkb->lkb_rqmode));
if (lkb->lkb_wait_type)
seq_printf(s, " wait_type: %d", lkb->lkb_wait_type);
- rv = seq_printf(s, "\n");
+ rv = seq_puts(s, "\n");
}
out:
unlock_rsb(res);
@@ -257,12 +257,12 @@ static int print_format3_lock(struct seq_file *s, struct dlm_lkb *lkb,
lkb->lkb_status,
lkb->lkb_grmode,
lkb->lkb_rqmode,
- lkb->lkb_bastmode,
+ lkb->lkb_last_bast.mode,
rsb_lookup,
lkb->lkb_wait_type,
lkb->lkb_lvbseq,
(unsigned long long)ktime_to_ns(lkb->lkb_timestamp),
- (unsigned long long)ktime_to_ns(lkb->lkb_time_bast));
+ (unsigned long long)ktime_to_ns(lkb->lkb_last_bast_time));
return rv;
}
@@ -300,7 +300,7 @@ static int print_format3(struct dlm_rsb *r, struct seq_file *s)
else
seq_printf(s, " %02x", (unsigned char)r->res_name[i]);
}
- rv = seq_printf(s, "\n");
+ rv = seq_puts(s, "\n");
if (rv)
goto out;
@@ -311,7 +311,7 @@ static int print_format3(struct dlm_rsb *r, struct seq_file *s)
for (i = 0; i < lvblen; i++)
seq_printf(s, " %02x", (unsigned char)r->res_lvbptr[i]);
- rv = seq_printf(s, "\n");
+ rv = seq_puts(s, "\n");
if (rv)
goto out;
@@ -344,6 +344,45 @@ static int print_format3(struct dlm_rsb *r, struct seq_file *s)
return rv;
}
+static int print_format4(struct dlm_rsb *r, struct seq_file *s)
+{
+ int our_nodeid = dlm_our_nodeid();
+ int print_name = 1;
+ int i, rv;
+
+ lock_rsb(r);
+
+ rv = seq_printf(s, "rsb %p %d %d %d %d %lu %lx %d ",
+ r,
+ r->res_nodeid,
+ r->res_master_nodeid,
+ r->res_dir_nodeid,
+ our_nodeid,
+ r->res_toss_time,
+ r->res_flags,
+ r->res_length);
+ if (rv)
+ goto out;
+
+ for (i = 0; i < r->res_length; i++) {
+ if (!isascii(r->res_name[i]) || !isprint(r->res_name[i]))
+ print_name = 0;
+ }
+
+ seq_printf(s, "%s", print_name ? "str " : "hex");
+
+ for (i = 0; i < r->res_length; i++) {
+ if (print_name)
+ seq_printf(s, "%c", r->res_name[i]);
+ else
+ seq_printf(s, " %02x", (unsigned char)r->res_name[i]);
+ }
+ rv = seq_puts(s, "\n");
+ out:
+ unlock_rsb(r);
+ return rv;
+}
+
struct rsbtbl_iter {
struct dlm_rsb *rsb;
unsigned bucket;
@@ -382,6 +421,13 @@ static int table_seq_show(struct seq_file *seq, void *iter_ptr)
}
rv = print_format3(ri->rsb, seq);
break;
+ case 4:
+ if (ri->header) {
+ seq_printf(seq, "version 4 rsb 2\n");
+ ri->header = 0;
+ }
+ rv = print_format4(ri->rsb, seq);
+ break;
}
return rv;
@@ -390,14 +436,18 @@ static int table_seq_show(struct seq_file *seq, void *iter_ptr)
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 void *table_seq_start(struct seq_file *seq, loff_t *pos)
{
+ struct rb_root *tree;
+ struct rb_node *node;
struct dlm_ls *ls = seq->private;
struct rsbtbl_iter *ri;
struct dlm_rsb *r;
loff_t n = *pos;
unsigned bucket, entry;
+ int toss = (seq->op == &format4_seq_ops);
bucket = n >> 32;
entry = n & ((1LL << 32) - 1);
@@ -416,11 +466,15 @@ static void *table_seq_start(struct seq_file *seq, loff_t *pos)
ri->format = 2;
if (seq->op == &format3_seq_ops)
ri->format = 3;
+ if (seq->op == &format4_seq_ops)
+ ri->format = 4;
+
+ tree = toss ? &ls->ls_rsbtbl[bucket].toss : &ls->ls_rsbtbl[bucket].keep;
spin_lock(&ls->ls_rsbtbl[bucket].lock);
- if (!list_empty(&ls->ls_rsbtbl[bucket].list)) {
- list_for_each_entry(r, &ls->ls_rsbtbl[bucket].list,
- res_hashchain) {
+ if (!RB_EMPTY_ROOT(tree)) {
+ for (node = rb_first(tree); node; node = rb_next(node)) {
+ r = rb_entry(node, struct dlm_rsb, res_hashnode);
if (!entry--) {
dlm_hold_rsb(r);
ri->rsb = r;
@@ -447,11 +501,12 @@ static void *table_seq_start(struct seq_file *seq, loff_t *pos)
kfree(ri);
return NULL;
}
+ tree = toss ? &ls->ls_rsbtbl[bucket].toss : &ls->ls_rsbtbl[bucket].keep;
spin_lock(&ls->ls_rsbtbl[bucket].lock);
- if (!list_empty(&ls->ls_rsbtbl[bucket].list)) {
- r = list_first_entry(&ls->ls_rsbtbl[bucket].list,
- struct dlm_rsb, res_hashchain);
+ if (!RB_EMPTY_ROOT(tree)) {
+ node = rb_first(tree);
+ r = rb_entry(node, struct dlm_rsb, res_hashnode);
dlm_hold_rsb(r);
ri->rsb = r;
ri->bucket = bucket;
@@ -467,10 +522,12 @@ static void *table_seq_next(struct seq_file *seq, void *iter_ptr, loff_t *pos)
{
struct dlm_ls *ls = seq->private;
struct rsbtbl_iter *ri = iter_ptr;
- struct list_head *next;
+ struct rb_root *tree;
+ struct rb_node *next;
struct dlm_rsb *r, *rp;
loff_t n = *pos;
unsigned bucket;
+ int toss = (seq->op == &format4_seq_ops);
bucket = n >> 32;
@@ -480,10 +537,10 @@ static void *table_seq_next(struct seq_file *seq, void *iter_ptr, loff_t *pos)
spin_lock(&ls->ls_rsbtbl[bucket].lock);
rp = ri->rsb;
- next = rp->res_hashchain.next;
+ next = rb_next(&rp->res_hashnode);
- if (next != &ls->ls_rsbtbl[bucket].list) {
- r = list_entry(next, struct dlm_rsb, res_hashchain);
+ if (next) {
+ r = rb_entry(next, struct dlm_rsb, res_hashnode);
dlm_hold_rsb(r);
ri->rsb = r;
spin_unlock(&ls->ls_rsbtbl[bucket].lock);
@@ -509,11 +566,12 @@ static void *table_seq_next(struct seq_file *seq, void *iter_ptr, loff_t *pos)
kfree(ri);
return NULL;
}
+ tree = toss ? &ls->ls_rsbtbl[bucket].toss : &ls->ls_rsbtbl[bucket].keep;
spin_lock(&ls->ls_rsbtbl[bucket].lock);
- if (!list_empty(&ls->ls_rsbtbl[bucket].list)) {
- r = list_first_entry(&ls->ls_rsbtbl[bucket].list,
- struct dlm_rsb, res_hashchain);
+ if (!RB_EMPTY_ROOT(tree)) {
+ next = rb_first(tree);
+ r = rb_entry(next, struct dlm_rsb, res_hashnode);
dlm_hold_rsb(r);
ri->rsb = r;
ri->bucket = bucket;
@@ -556,9 +614,17 @@ static const struct seq_operations format3_seq_ops = {
.show = table_seq_show,
};
+static const struct seq_operations format4_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 int table_open(struct inode *inode, struct file *file)
{
@@ -571,6 +637,8 @@ static int table_open(struct inode *inode, struct file *file)
ret = seq_open(file, &format2_seq_ops);
else if (file->f_op == &format3_fops)
ret = seq_open(file, &format3_seq_ops);
+ else if (file->f_op == &format4_fops)
+ ret = seq_open(file, &format4_seq_ops);
if (ret)
return ret;
@@ -604,16 +672,17 @@ static const struct file_operations format3_fops = {
.release = seq_release
};
+static const struct file_operations format4_fops = {
+ .owner = THIS_MODULE,
+ .open = table_open,
+ .read = seq_read,
+ .llseek = seq_lseek,
+ .release = seq_release
+};
+
/*
* dump lkb's on the ls_waiters list
*/
-
-static int waiters_open(struct inode *inode, struct file *file)
-{
- file->private_data = inode->i_private;
- return 0;
-}
-
static ssize_t waiters_read(struct file *file, char __user *userbuf,
size_t count, loff_t *ppos)
{
@@ -642,7 +711,7 @@ static ssize_t waiters_read(struct file *file, char __user *userbuf,
static const struct file_operations waiters_fops = {
.owner = THIS_MODULE,
- .open = waiters_open,
+ .open = simple_open,
.read = waiters_read,
.llseek = default_llseek,
};
@@ -657,6 +726,8 @@ void dlm_delete_debug_file(struct dlm_ls *ls)
debugfs_remove(ls->ls_debug_locks_dentry);
if (ls->ls_debug_all_dentry)
debugfs_remove(ls->ls_debug_all_dentry);
+ if (ls->ls_debug_toss_dentry)
+ debugfs_remove(ls->ls_debug_toss_dentry);
}
int dlm_create_debug_file(struct dlm_ls *ls)
@@ -699,6 +770,19 @@ int dlm_create_debug_file(struct dlm_ls *ls)
if (!ls->ls_debug_all_dentry)
goto fail;
+ /* format 4 */
+
+ memset(name, 0, sizeof(name));
+ snprintf(name, DLM_LOCKSPACE_LEN+8, "%s_toss", ls->ls_name);
+
+ ls->ls_debug_toss_dentry = debugfs_create_file(name,
+ S_IFREG | S_IRUGO,
+ dlm_root,
+ ls,
+ &format4_fops);
+ if (!ls->ls_debug_toss_dentry)
+ goto fail;
+
memset(name, 0, sizeof(name));
snprintf(name, DLM_LOCKSPACE_LEN+8, "%s_waiters", ls->ls_name);
diff --git a/fs/dlm/dir.c b/fs/dlm/dir.c
index 7b84c1dbc82..d975851a7e1 100644
--- a/fs/dlm/dir.c
+++ b/fs/dlm/dir.c
@@ -23,50 +23,6 @@
#include "lock.h"
#include "dir.h"
-
-static void put_free_de(struct dlm_ls *ls, struct dlm_direntry *de)
-{
- spin_lock(&ls->ls_recover_list_lock);
- list_add(&de->list, &ls->ls_recover_list);
- spin_unlock(&ls->ls_recover_list_lock);
-}
-
-static struct dlm_direntry *get_free_de(struct dlm_ls *ls, int len)
-{
- int found = 0;
- struct dlm_direntry *de;
-
- spin_lock(&ls->ls_recover_list_lock);
- list_for_each_entry(de, &ls->ls_recover_list, list) {
- if (de->length == len) {
- list_del(&de->list);
- de->master_nodeid = 0;
- memset(de->name, 0, len);
- found = 1;
- break;
- }
- }
- spin_unlock(&ls->ls_recover_list_lock);
-
- if (!found)
- de = kzalloc(sizeof(struct dlm_direntry) + len, GFP_NOFS);
- return de;
-}
-
-void dlm_clear_free_entries(struct dlm_ls *ls)
-{
- struct dlm_direntry *de;
-
- spin_lock(&ls->ls_recover_list_lock);
- while (!list_empty(&ls->ls_recover_list)) {
- de = list_entry(ls->ls_recover_list.next, struct dlm_direntry,
- list);
- list_del(&de->list);
- kfree(de);
- }
- spin_unlock(&ls->ls_recover_list_lock);
-}
-
/*
* We use the upper 16 bits of the hash value to select the directory node.
* Low bits are used for distribution of rsb's among hash buckets on each node.
@@ -78,144 +34,53 @@ void dlm_clear_free_entries(struct dlm_ls *ls)
int dlm_hash2nodeid(struct dlm_ls *ls, uint32_t hash)
{
- struct list_head *tmp;
- struct dlm_member *memb = NULL;
- uint32_t node, n = 0;
- int nodeid;
-
- if (ls->ls_num_nodes == 1) {
- nodeid = dlm_our_nodeid();
- goto out;
- }
+ uint32_t node;
- if (ls->ls_node_array) {
+ if (ls->ls_num_nodes == 1)
+ return dlm_our_nodeid();
+ else {
node = (hash >> 16) % ls->ls_total_weight;
- nodeid = ls->ls_node_array[node];
- goto out;
- }
-
- /* make_member_array() failed to kmalloc ls_node_array... */
-
- node = (hash >> 16) % ls->ls_num_nodes;
-
- list_for_each(tmp, &ls->ls_nodes) {
- if (n++ != node)
- continue;
- memb = list_entry(tmp, struct dlm_member, list);
- break;
+ return ls->ls_node_array[node];
}
-
- DLM_ASSERT(memb , printk("num_nodes=%u n=%u node=%u\n",
- ls->ls_num_nodes, n, node););
- nodeid = memb->nodeid;
- out:
- return nodeid;
}
int dlm_dir_nodeid(struct dlm_rsb *r)
{
- return dlm_hash2nodeid(r->res_ls, r->res_hash);
-}
-
-static inline uint32_t dir_hash(struct dlm_ls *ls, char *name, int len)
-{
- uint32_t val;
-
- val = jhash(name, len, 0);
- val &= (ls->ls_dirtbl_size - 1);
-
- return val;
-}
-
-static void add_entry_to_hash(struct dlm_ls *ls, struct dlm_direntry *de)
-{
- uint32_t bucket;
-
- bucket = dir_hash(ls, de->name, de->length);
- list_add_tail(&de->list, &ls->ls_dirtbl[bucket].list);
-}
-
-static struct dlm_direntry *search_bucket(struct dlm_ls *ls, char *name,
- int namelen, uint32_t bucket)
-{
- struct dlm_direntry *de;
-
- list_for_each_entry(de, &ls->ls_dirtbl[bucket].list, list) {
- if (de->length == namelen && !memcmp(name, de->name, namelen))
- goto out;
- }
- de = NULL;
- out:
- return de;
+ return r->res_dir_nodeid;
}
-void dlm_dir_remove_entry(struct dlm_ls *ls, int nodeid, char *name, int namelen)
+void dlm_recover_dir_nodeid(struct dlm_ls *ls)
{
- struct dlm_direntry *de;
- uint32_t bucket;
-
- bucket = dir_hash(ls, name, namelen);
-
- spin_lock(&ls->ls_dirtbl[bucket].lock);
-
- de = search_bucket(ls, name, namelen, bucket);
-
- if (!de) {
- log_error(ls, "remove fr %u none", nodeid);
- goto out;
- }
-
- if (de->master_nodeid != nodeid) {
- log_error(ls, "remove fr %u ID %u", nodeid, de->master_nodeid);
- goto out;
- }
-
- list_del(&de->list);
- kfree(de);
- out:
- spin_unlock(&ls->ls_dirtbl[bucket].lock);
-}
+ struct dlm_rsb *r;
-void dlm_dir_clear(struct dlm_ls *ls)
-{
- struct list_head *head;
- struct dlm_direntry *de;
- int i;
-
- DLM_ASSERT(list_empty(&ls->ls_recover_list), );
-
- for (i = 0; i < ls->ls_dirtbl_size; i++) {
- spin_lock(&ls->ls_dirtbl[i].lock);
- head = &ls->ls_dirtbl[i].list;
- while (!list_empty(head)) {
- de = list_entry(head->next, struct dlm_direntry, list);
- list_del(&de->list);
- put_free_de(ls, de);
- }
- spin_unlock(&ls->ls_dirtbl[i].lock);
+ down_read(&ls->ls_root_sem);
+ list_for_each_entry(r, &ls->ls_root_list, res_root_list) {
+ r->res_dir_nodeid = dlm_hash2nodeid(ls, r->res_hash);
}
+ up_read(&ls->ls_root_sem);
}
int dlm_recover_directory(struct dlm_ls *ls)
{
struct dlm_member *memb;
- struct dlm_direntry *de;
char *b, *last_name = NULL;
- int error = -ENOMEM, last_len, count = 0;
+ int error = -ENOMEM, last_len, nodeid, result;
uint16_t namelen;
+ unsigned int count = 0, count_match = 0, count_bad = 0, count_add = 0;
- log_debug(ls, "dlm_recover_directory");
+ log_rinfo(ls, "dlm_recover_directory");
if (dlm_no_directory(ls))
goto out_status;
- dlm_dir_clear(ls);
-
last_name = kmalloc(DLM_RESNAME_MAXLEN, GFP_NOFS);
if (!last_name)
goto out;
list_for_each_entry(memb, &ls->ls_nodes, list) {
+ if (memb->nodeid == dlm_our_nodeid())
+ continue;
+
memset(last_name, 0, DLM_RESNAME_MAXLEN);
last_len = 0;
@@ -230,7 +95,7 @@ int dlm_recover_directory(struct dlm_ls *ls)
if (error)
goto out_free;
- schedule();
+ cond_resched();
/*
* pick namelen/name pairs out of received buffer
@@ -267,96 +132,96 @@ int dlm_recover_directory(struct dlm_ls *ls)
if (namelen > DLM_RESNAME_MAXLEN)
goto out_free;
- error = -ENOMEM;
- de = get_free_de(ls, namelen);
- if (!de)
+ error = dlm_master_lookup(ls, memb->nodeid,
+ b, namelen,
+ DLM_LU_RECOVER_DIR,
+ &nodeid, &result);
+ if (error) {
+ log_error(ls, "recover_dir lookup %d",
+ error);
goto out_free;
+ }
+
+ /* The name was found in rsbtbl, but the
+ * master nodeid is different from
+ * memb->nodeid which says it is the master.
+ * This should not happen. */
+
+ if (result == DLM_LU_MATCH &&
+ nodeid != memb->nodeid) {
+ count_bad++;
+ log_error(ls, "recover_dir lookup %d "
+ "nodeid %d memb %d bad %u",
+ result, nodeid, memb->nodeid,
+ count_bad);
+ print_hex_dump_bytes("dlm_recover_dir ",
+ DUMP_PREFIX_NONE,
+ b, namelen);
+ }
+
+ /* The name was found in rsbtbl, and the
+ * master nodeid matches memb->nodeid. */
+
+ if (result == DLM_LU_MATCH &&
+ nodeid == memb->nodeid) {
+ count_match++;
+ }
+
+ /* The name was not found in rsbtbl and was
+ * added with memb->nodeid as the master. */
+
+ if (result == DLM_LU_ADD) {
+ count_add++;
+ }
- de->master_nodeid = memb->nodeid;
- de->length = namelen;
last_len = namelen;
- memcpy(de->name, b, namelen);
memcpy(last_name, b, namelen);
b += namelen;
left -= namelen;
-
- add_entry_to_hash(ls, de);
count++;
}
}
- done:
+ done:
;
}
out_status:
error = 0;
dlm_set_recover_status(ls, DLM_RS_DIR);
- log_debug(ls, "dlm_recover_directory %d entries", count);
+
+ log_rinfo(ls, "dlm_recover_directory %u in %u new",
+ count, count_add);
out_free:
kfree(last_name);
out:
- dlm_clear_free_entries(ls);
return error;
}
-static int get_entry(struct dlm_ls *ls, int nodeid, char *name,
- int namelen, int *r_nodeid)
+static struct dlm_rsb *find_rsb_root(struct dlm_ls *ls, char *name, int len)
{
- struct dlm_direntry *de, *tmp;
- uint32_t bucket;
-
- bucket = dir_hash(ls, name, namelen);
-
- spin_lock(&ls->ls_dirtbl[bucket].lock);
- de = search_bucket(ls, name, namelen, bucket);
- if (de) {
- *r_nodeid = de->master_nodeid;
- spin_unlock(&ls->ls_dirtbl[bucket].lock);
- if (*r_nodeid == nodeid)
- return -EEXIST;
- return 0;
- }
-
- spin_unlock(&ls->ls_dirtbl[bucket].lock);
-
- if (namelen > DLM_RESNAME_MAXLEN)
- return -EINVAL;
-
- de = kzalloc(sizeof(struct dlm_direntry) + namelen, GFP_NOFS);
- if (!de)
- return -ENOMEM;
-
- de->master_nodeid = nodeid;
- de->length = namelen;
- memcpy(de->name, name, namelen);
+ struct dlm_rsb *r;
+ uint32_t hash, bucket;
+ int rv;
- spin_lock(&ls->ls_dirtbl[bucket].lock);
- tmp = search_bucket(ls, name, namelen, bucket);
- if (tmp) {
- kfree(de);
- de = tmp;
- } else {
- list_add_tail(&de->list, &ls->ls_dirtbl[bucket].list);
- }
- *r_nodeid = de->master_nodeid;
- spin_unlock(&ls->ls_dirtbl[bucket].lock);
- return 0;
-}
+ hash = jhash(name, len, 0);
+ bucket = hash & (ls->ls_rsbtbl_size - 1);
-int dlm_dir_lookup(struct dlm_ls *ls, int nodeid, char *name, int namelen,
- int *r_nodeid)
-{
- return get_entry(ls, nodeid, name, namelen, r_nodeid);
-}
+ spin_lock(&ls->ls_rsbtbl[bucket].lock);
+ rv = dlm_search_rsb_tree(&ls->ls_rsbtbl[bucket].keep, name, len, &r);
+ if (rv)
+ rv = dlm_search_rsb_tree(&ls->ls_rsbtbl[bucket].toss,
+ name, len, &r);
+ spin_unlock(&ls->ls_rsbtbl[bucket].lock);
-static struct dlm_rsb *find_rsb_root(struct dlm_ls *ls, char *name, int len)
-{
- struct dlm_rsb *r;
+ if (!rv)
+ return r;
down_read(&ls->ls_root_sem);
list_for_each_entry(r, &ls->ls_root_list, res_root_list) {
if (len == r->res_length && !memcmp(name, r->res_name, len)) {
up_read(&ls->ls_root_sem);
+ log_debug(ls, "find_rsb_root revert to root_list %s",
+ r->res_name);
return r;
}
}
@@ -413,6 +278,7 @@ void dlm_copy_master_names(struct dlm_ls *ls, char *inbuf, int inlen,
be_namelen = cpu_to_be16(0);
memcpy(outbuf + offset, &be_namelen, sizeof(__be16));
offset += sizeof(__be16);
+ ls->ls_recover_dir_sent_msg++;
goto out;
}
@@ -421,6 +287,7 @@ void dlm_copy_master_names(struct dlm_ls *ls, char *inbuf, int inlen,
offset += sizeof(__be16);
memcpy(outbuf + offset, r->res_name, r->res_length);
offset += r->res_length;
+ ls->ls_recover_dir_sent_res++;
}
/*
@@ -433,8 +300,8 @@ void dlm_copy_master_names(struct dlm_ls *ls, char *inbuf, int inlen,
be_namelen = cpu_to_be16(0xFFFF);
memcpy(outbuf + offset, &be_namelen, sizeof(__be16));
offset += sizeof(__be16);
+ ls->ls_recover_dir_sent_msg++;
}
-
out:
up_read(&ls->ls_root_sem);
}
diff --git a/fs/dlm/dir.h b/fs/dlm/dir.h
index 0b0eb1267b6..41750634445 100644
--- a/fs/dlm/dir.h
+++ b/fs/dlm/dir.h
@@ -14,15 +14,10 @@
#ifndef __DIR_DOT_H__
#define __DIR_DOT_H__
-
int dlm_dir_nodeid(struct dlm_rsb *rsb);
int dlm_hash2nodeid(struct dlm_ls *ls, uint32_t hash);
-void dlm_dir_remove_entry(struct dlm_ls *ls, int nodeid, char *name, int len);
-void dlm_dir_clear(struct dlm_ls *ls);
-void dlm_clear_free_entries(struct dlm_ls *ls);
+void dlm_recover_dir_nodeid(struct dlm_ls *ls);
int dlm_recover_directory(struct dlm_ls *ls);
-int dlm_dir_lookup(struct dlm_ls *ls, int nodeid, char *name, int namelen,
- int *r_nodeid);
void dlm_copy_master_names(struct dlm_ls *ls, char *inbuf, int inlen,
char *outbuf, int outlen, int nodeid);
diff --git a/fs/dlm/dlm_internal.h b/fs/dlm/dlm_internal.h
index f632b58cd22..5eff6ea3e27 100644
--- a/fs/dlm/dlm_internal.h
+++ b/fs/dlm/dlm_internal.h
@@ -2,7 +2,7 @@
*******************************************************************************
**
** Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
-** Copyright (C) 2004-2010 Red Hat, Inc. All rights reserved.
+** Copyright (C) 2004-2011 Red Hat, Inc. All rights reserved.
**
** This copyrighted material is made available to anyone wishing to use,
** modify, copy, or redistribute it subject to the terms and conditions
@@ -37,6 +37,8 @@
#include <linux/jhash.h>
#include <linux/miscdevice.h>
#include <linux/mutex.h>
+#include <linux/idr.h>
+#include <linux/ratelimit.h>
#include <asm/uaccess.h>
#include <linux/dlm.h>
@@ -52,10 +54,7 @@ struct dlm_ls;
struct dlm_lkb;
struct dlm_rsb;
struct dlm_member;
-struct dlm_lkbtable;
struct dlm_rsbtable;
-struct dlm_dirtable;
-struct dlm_direntry;
struct dlm_recover;
struct dlm_header;
struct dlm_message;
@@ -66,6 +65,8 @@ struct dlm_mhandle;
printk(KERN_ERR "dlm: "fmt"\n" , ##args)
#define log_error(ls, fmt, args...) \
printk(KERN_ERR "dlm: %s: " fmt "\n", (ls)->ls_name , ##args)
+#define log_rinfo(ls, fmt, args...) \
+ printk(KERN_INFO "dlm: %s: " fmt "\n", (ls)->ls_name , ##args);
#define log_debug(ls, fmt, args...) \
do { \
@@ -74,6 +75,13 @@ do { \
(ls)->ls_name , ##args); \
} while (0)
+#define log_limit(ls, fmt, args...) \
+do { \
+ if (dlm_config.ci_log_debug) \
+ printk_ratelimited(KERN_DEBUG "dlm: %s: " fmt "\n", \
+ (ls)->ls_name , ##args); \
+} while (0)
+
#define DLM_ASSERT(x, do) \
{ \
if (!(x)) \
@@ -90,29 +98,15 @@ do { \
}
-struct dlm_direntry {
- struct list_head list;
- uint32_t master_nodeid;
- uint16_t length;
- char name[1];
-};
-
-struct dlm_dirtable {
- struct list_head list;
- spinlock_t lock;
-};
+#define DLM_RTF_SHRINK 0x00000001
struct dlm_rsbtable {
- struct list_head list;
- struct list_head toss;
+ struct rb_root keep;
+ struct rb_root toss;
spinlock_t lock;
+ uint32_t flags;
};
-struct dlm_lkbtable {
- struct list_head list;
- rwlock_t lock;
- uint16_t counter;
-};
/*
* Lockspace member (per node in a ls)
@@ -122,6 +116,10 @@ struct dlm_member {
struct list_head list;
int nodeid;
int weight;
+ int slot;
+ int slot_prev;
+ int comm_seq;
+ uint32_t generation;
};
/*
@@ -130,10 +128,8 @@ struct dlm_member {
struct dlm_recover {
struct list_head list;
- int *nodeids; /* nodeids of all members */
- int node_count;
- int *new; /* nodeids of new members */
- int new_count;
+ struct dlm_config_node *nodes;
+ int nodes_count;
uint64_t seq;
};
@@ -192,11 +188,6 @@ struct dlm_args {
* lkb is a process copy, the nodeid specifies the lock master.
*/
-/* lkb_ast_type */
-
-#define AST_COMP 1
-#define AST_BAST 2
-
/* lkb_status */
#define DLM_LKSTS_WAITING 1
@@ -214,9 +205,24 @@ struct dlm_args {
#define DLM_IFL_WATCH_TIMEWARN 0x00400000
#define DLM_IFL_TIMEOUT_CANCEL 0x00800000
#define DLM_IFL_DEADLOCK_CANCEL 0x01000000
+#define DLM_IFL_STUB_MS 0x02000000 /* magic number for m_flags */
#define DLM_IFL_USER 0x00000001
#define DLM_IFL_ORPHAN 0x00000002
+#define DLM_CALLBACKS_SIZE 6
+
+#define DLM_CB_CAST 0x00000001
+#define DLM_CB_BAST 0x00000002
+#define DLM_CB_SKIP 0x00000004
+
+struct dlm_callback {
+ uint64_t seq;
+ uint32_t flags; /* DLM_CBF_ */
+ int sb_status; /* copy to lksb status */
+ uint8_t sb_flags; /* copy to lksb flags */
+ int8_t mode; /* rq mode of bast, gr mode of cast */
+};
+
struct dlm_lkb {
struct dlm_rsb *lkb_resource; /* the rsb */
struct kref lkb_ref;
@@ -236,25 +242,28 @@ struct dlm_lkb {
int8_t lkb_wait_type; /* type of reply waiting for */
int8_t lkb_wait_count;
- int8_t lkb_ast_type; /* type of ast queued for */
- int8_t lkb_ast_first; /* type of first ast queued */
-
- int8_t lkb_bastmode; /* req mode of queued bast */
- int8_t lkb_castmode; /* gr mode of queued cast */
- int8_t lkb_bastmode_done; /* last delivered bastmode */
- int8_t lkb_castmode_done; /* last delivered castmode */
+ int lkb_wait_nodeid; /* for debugging */
- struct list_head lkb_idtbl_list; /* lockspace lkbtbl */
struct list_head lkb_statequeue; /* rsb g/c/w list */
struct list_head lkb_rsb_lookup; /* waiting for rsb lookup */
struct list_head lkb_wait_reply; /* waiting for remote reply */
- struct list_head lkb_astqueue; /* need ast to be sent */
struct list_head lkb_ownqueue; /* list of locks for a process */
struct list_head lkb_time_list;
- ktime_t lkb_time_bast; /* for debugging */
ktime_t lkb_timestamp;
+ ktime_t lkb_wait_time;
unsigned long lkb_timeout_cs;
+ struct mutex lkb_cb_mutex;
+ struct work_struct lkb_cb_work;
+ struct list_head lkb_cb_list; /* for ls_cb_delay or proc->asts */
+ struct dlm_callback lkb_callbacks[DLM_CALLBACKS_SIZE];
+ struct dlm_callback lkb_last_cast;
+ struct dlm_callback lkb_last_bast;
+ ktime_t lkb_last_cast_time; /* for debugging */
+ ktime_t lkb_last_bast_time; /* for debugging */
+
+ uint64_t lkb_recover_seq; /* from ls_recover_seq */
+
char *lkb_lvbptr;
struct dlm_lksb *lkb_lksb; /* caller's status block */
void (*lkb_astfn) (void *astparam);
@@ -265,6 +274,15 @@ struct dlm_lkb {
};
};
+/*
+ * res_master_nodeid is "normal": 0 is unset/invalid, non-zero is the real
+ * nodeid, even when nodeid is our_nodeid.
+ *
+ * res_nodeid is "odd": -1 is unset/invalid, zero means our_nodeid,
+ * greater than zero when another nodeid.
+ *
+ * (TODO: remove res_nodeid and only use res_master_nodeid)
+ */
struct dlm_rsb {
struct dlm_ls *res_ls; /* the lockspace */
@@ -273,13 +291,19 @@ struct dlm_rsb {
unsigned long res_flags;
int res_length; /* length of rsb name */
int res_nodeid;
+ int res_master_nodeid;
+ int res_dir_nodeid;
+ int res_id; /* for ls_recover_idr */
uint32_t res_lvbseq;
uint32_t res_hash;
uint32_t res_bucket; /* rsbtbl */
unsigned long res_toss_time;
uint32_t res_first_lkid;
struct list_head res_lookup; /* lkbs waiting on first */
- struct list_head res_hashchain; /* rsbtbl */
+ union {
+ struct list_head res_hashchain;
+ struct rb_node res_hashnode; /* rsbtbl */
+ };
struct list_head res_grantqueue;
struct list_head res_convertqueue;
struct list_head res_waitqueue;
@@ -289,13 +313,24 @@ struct dlm_rsb {
int res_recover_locks_count;
char *res_lvbptr;
- char res_name[1];
+ char res_name[DLM_RESNAME_MAXLEN+1];
};
+/* dlm_master_lookup() flags */
+
+#define DLM_LU_RECOVER_DIR 1
+#define DLM_LU_RECOVER_MASTER 2
+
+/* dlm_master_lookup() results */
+
+#define DLM_LU_MATCH 1
+#define DLM_LU_ADD 2
+
/* find_rsb() flags */
-#define R_MASTER 1 /* only return rsb if it's a master */
-#define R_CREATE 2 /* create/add rsb if not found */
+#define R_REQUEST 0x00000001
+#define R_RECEIVE_REQUEST 0x00000002
+#define R_RECEIVE_RECOVER 0x00000004
/* rsb_flags */
@@ -306,7 +341,8 @@ enum rsb_flags {
RSB_NEW_MASTER,
RSB_NEW_MASTER2,
RSB_RECOVER_CONVERT,
- RSB_LOCKS_PURGED,
+ RSB_RECOVER_GRANT,
+ RSB_RECOVER_LVB_INVAL,
};
static inline void rsb_set_flag(struct dlm_rsb *r, enum rsb_flags flag)
@@ -328,7 +364,9 @@ static inline int rsb_flag(struct dlm_rsb *r, enum rsb_flags flag)
/* dlm_header is first element of all structs sent between nodes */
#define DLM_HEADER_MAJOR 0x00030000
-#define DLM_HEADER_MINOR 0x00000000
+#define DLM_HEADER_MINOR 0x00000001
+
+#define DLM_HEADER_SLOTS 0x00000001
#define DLM_MSG 1
#define DLM_RCOM 2
@@ -416,10 +454,34 @@ union dlm_packet {
struct dlm_rcom rcom;
};
+#define DLM_RSF_NEED_SLOTS 0x00000001
+
+/* RCOM_STATUS data */
+struct rcom_status {
+ __le32 rs_flags;
+ __le32 rs_unused1;
+ __le64 rs_unused2;
+};
+
+/* RCOM_STATUS_REPLY data */
struct rcom_config {
__le32 rf_lvblen;
__le32 rf_lsflags;
- __le64 rf_unused;
+
+ /* DLM_HEADER_SLOTS adds: */
+ __le32 rf_flags;
+ __le16 rf_our_slot;
+ __le16 rf_num_slots;
+ __le32 rf_generation;
+ __le32 rf_unused1;
+ __le64 rf_unused2;
+};
+
+struct rcom_slot {
+ __le32 ro_nodeid;
+ __le16 ro_slot;
+ __le16 ro_unused1;
+ __le64 ro_unused2;
};
struct rcom_lock {
@@ -442,10 +504,18 @@ struct rcom_lock {
char rl_lvb[0];
};
+/*
+ * The max number of resources per rsbtbl bucket that shrink will attempt
+ * to remove in each iteration.
+ */
+
+#define DLM_REMOVE_NAMES_MAX 8
+
struct dlm_ls {
struct list_head ls_list; /* list of lockspaces */
dlm_lockspace_t *ls_local_handle;
uint32_t ls_global_id; /* global unique lockspace ID */
+ uint32_t ls_generation;
uint32_t ls_exflags;
int ls_lvblen;
int ls_count; /* refcount of processes in
@@ -455,15 +525,12 @@ struct dlm_ls {
unsigned long ls_scan_time;
struct kobject ls_kobj;
+ struct idr ls_lkbidr;
+ spinlock_t ls_lkbidr_spin;
+
struct dlm_rsbtable *ls_rsbtbl;
uint32_t ls_rsbtbl_size;
- struct dlm_lkbtable *ls_lkbtbl;
- uint32_t ls_lkbtbl_size;
-
- struct dlm_dirtable *ls_dirtbl;
- uint32_t ls_dirtbl_size;
-
struct mutex ls_waiters_mutex;
struct list_head ls_waiters; /* lkbs needing a reply */
@@ -473,6 +540,16 @@ struct dlm_ls {
struct mutex ls_timeout_mutex;
struct list_head ls_timeout;
+ spinlock_t ls_new_rsb_spin;
+ int ls_new_rsb_count;
+ struct list_head ls_new_rsb; /* new rsb structs */
+
+ spinlock_t ls_remove_spin;
+ char ls_remove_name[DLM_RESNAME_MAXLEN+1];
+ char *ls_remove_names[DLM_REMOVE_NAMES_MAX];
+ int ls_remove_len;
+ int ls_remove_lens[DLM_REMOVE_NAMES_MAX];
+
struct list_head ls_nodes; /* current nodes in ls */
struct list_head ls_nodes_gone; /* dead node list, recovery */
int ls_num_nodes; /* number of nodes in ls */
@@ -480,6 +557,11 @@ struct dlm_ls {
int ls_total_weight;
int *ls_node_array;
+ int ls_slot;
+ int ls_num_slots;
+ int ls_slots_size;
+ struct dlm_slot *ls_slots;
+
struct dlm_rsb ls_stub_rsb; /* for returning errors */
struct dlm_lkb ls_stub_lkb; /* for returning errors */
struct dlm_message ls_stub_ms; /* for faking a reply */
@@ -488,6 +570,7 @@ struct dlm_ls {
struct dentry *ls_debug_waiters_dentry; /* debugfs */
struct dentry *ls_debug_locks_dentry; /* debugfs */
struct dentry *ls_debug_all_dentry; /* debugfs */
+ struct dentry *ls_debug_toss_dentry; /* debugfs */
wait_queue_head_t ls_uevent_wait; /* user part of join/leave */
int ls_uevent_result;
@@ -496,8 +579,12 @@ struct dlm_ls {
struct miscdevice ls_device;
+ struct workqueue_struct *ls_callback_wq;
+
/* recovery related */
+ struct mutex ls_cb_mutex;
+ struct list_head ls_cb_delay; /* save for queue_work later */
struct timer_list ls_timer;
struct task_struct *ls_recoverd_task;
struct mutex ls_recoverd_active;
@@ -512,28 +599,64 @@ struct dlm_ls {
struct mutex ls_requestqueue_mutex;
struct dlm_rcom *ls_recover_buf;
int ls_recover_nodeid; /* for debugging */
+ unsigned int ls_recover_dir_sent_res; /* for log info */
+ unsigned int ls_recover_dir_sent_msg; /* for log info */
+ unsigned int ls_recover_locks_in; /* for log info */
uint64_t ls_rcom_seq;
spinlock_t ls_rcom_spin;
struct list_head ls_recover_list;
spinlock_t ls_recover_list_lock;
int ls_recover_list_count;
+ struct idr ls_recover_idr;
+ spinlock_t ls_recover_idr_lock;
wait_queue_head_t ls_wait_general;
+ wait_queue_head_t ls_recover_lock_wait;
struct mutex ls_clear_proc_locks;
struct list_head ls_root_list; /* root resources */
struct rw_semaphore ls_root_sem; /* protect root_list */
+ const struct dlm_lockspace_ops *ls_ops;
+ void *ls_ops_arg;
+
int ls_namelen;
char ls_name[1];
};
-#define LSFL_WORK 0
-#define LSFL_RUNNING 1
-#define LSFL_RECOVERY_STOP 2
-#define LSFL_RCOM_READY 3
-#define LSFL_RCOM_WAIT 4
-#define LSFL_UEVENT_WAIT 5
-#define LSFL_TIMEWARN 6
+/*
+ * LSFL_RECOVER_STOP - dlm_ls_stop() sets this to tell dlm recovery routines
+ * that they should abort what they're doing so new recovery can be started.
+ *
+ * LSFL_RECOVER_DOWN - dlm_ls_stop() sets this to tell dlm_recoverd that it
+ * should do down_write() on the in_recovery rw_semaphore. (doing down_write
+ * within dlm_ls_stop causes complaints about the lock acquired/released
+ * in different contexts.)
+ *
+ * LSFL_RECOVER_LOCK - dlm_recoverd holds the in_recovery rw_semaphore.
+ * It sets this after it is done with down_write() on the in_recovery
+ * rw_semaphore and clears it after it has released the rw_semaphore.
+ *
+ * LSFL_RECOVER_WORK - dlm_ls_start() sets this to tell dlm_recoverd that it
+ * should begin recovery of the lockspace.
+ *
+ * LSFL_RUNNING - set when normal locking activity is enabled.
+ * dlm_ls_stop() clears this to tell dlm locking routines that they should
+ * quit what they are doing so recovery can run. dlm_recoverd sets
+ * this after recovery is finished.
+ */
+
+#define LSFL_RECOVER_STOP 0
+#define LSFL_RECOVER_DOWN 1
+#define LSFL_RECOVER_LOCK 2
+#define LSFL_RECOVER_WORK 3
+#define LSFL_RUNNING 4
+
+#define LSFL_RCOM_READY 5
+#define LSFL_RCOM_WAIT 6
+#define LSFL_UEVENT_WAIT 7
+#define LSFL_TIMEWARN 8
+#define LSFL_CB_DELAY 9
+#define LSFL_NODIR 10
/* much of this is just saving user space pointers associated with the
lock that we pass back to the user lib with an ast */
@@ -544,8 +667,6 @@ struct dlm_user_args {
(dlm_user_proc) on the struct file,
the process's locks point back to it*/
struct dlm_lksb lksb;
- int old_mode;
- int update_user_lvb;
struct dlm_lksb __user *user_lksb;
void __user *castparam;
void __user *castaddr;
@@ -578,12 +699,12 @@ static inline int dlm_locking_stopped(struct dlm_ls *ls)
static inline int dlm_recovery_stopped(struct dlm_ls *ls)
{
- return test_bit(LSFL_RECOVERY_STOP, &ls->ls_flags);
+ return test_bit(LSFL_RECOVER_STOP, &ls->ls_flags);
}
static inline int dlm_no_directory(struct dlm_ls *ls)
{
- return (ls->ls_exflags & DLM_LSFL_NODIR) ? 1 : 0;
+ return test_bit(LSFL_NODIR, &ls->ls_flags);
}
int dlm_netlink_init(void);
diff --git a/fs/dlm/lock.c b/fs/dlm/lock.c
index 64e5f3efdd8..83f3d552030 100644
--- a/fs/dlm/lock.c
+++ b/fs/dlm/lock.c
@@ -56,6 +56,7 @@
L: receive_xxxx_reply() <- R: send_xxxx_reply()
*/
#include <linux/types.h>
+#include <linux/rbtree.h>
#include <linux/slab.h>
#include "dlm_internal.h"
#include <linux/dlm_device.h>
@@ -89,6 +90,7 @@ static void __receive_convert_reply(struct dlm_rsb *r, struct dlm_lkb *lkb,
static int receive_extralen(struct dlm_message *ms);
static void do_purge(struct dlm_ls *ls, int nodeid, int pid);
static void del_timeout(struct dlm_lkb *lkb);
+static void toss_rsb(struct kref *kref);
/*
* Lock compatibilty matrix - thanks Steve
@@ -159,18 +161,21 @@ static const int __quecvt_compat_matrix[8][8] = {
void dlm_print_lkb(struct dlm_lkb *lkb)
{
- printk(KERN_ERR "lkb: nodeid %d id %x remid %x exflags %x flags %x\n"
- " status %d rqmode %d grmode %d wait_type %d ast_type %d\n",
+ printk(KERN_ERR "lkb: nodeid %d id %x remid %x exflags %x flags %x "
+ "sts %d rq %d gr %d wait_type %d wait_nodeid %d seq %llu\n",
lkb->lkb_nodeid, lkb->lkb_id, lkb->lkb_remid, lkb->lkb_exflags,
lkb->lkb_flags, lkb->lkb_status, lkb->lkb_rqmode,
- lkb->lkb_grmode, lkb->lkb_wait_type, lkb->lkb_ast_type);
+ lkb->lkb_grmode, lkb->lkb_wait_type, lkb->lkb_wait_nodeid,
+ (unsigned long long)lkb->lkb_recover_seq);
}
static void dlm_print_rsb(struct dlm_rsb *r)
{
- printk(KERN_ERR "rsb: nodeid %d flags %lx first %x rlc %d name %s\n",
- r->res_nodeid, r->res_flags, r->res_first_lkid,
- r->res_recover_locks_count, r->res_name);
+ printk(KERN_ERR "rsb: nodeid %d master %d dir %d flags %lx first %x "
+ "rlc %d name %s\n",
+ r->res_nodeid, r->res_master_nodeid, r->res_dir_nodeid,
+ r->res_flags, r->res_first_lkid, r->res_recover_locks_count,
+ r->res_name);
}
void dlm_dump_rsb(struct dlm_rsb *r)
@@ -250,8 +255,6 @@ static inline int is_process_copy(struct dlm_lkb *lkb)
static inline int is_master_copy(struct dlm_lkb *lkb)
{
- if (lkb->lkb_flags & DLM_IFL_MSTCPY)
- DLM_ASSERT(lkb->lkb_nodeid, dlm_print_lkb(lkb););
return (lkb->lkb_flags & DLM_IFL_MSTCPY) ? 1 : 0;
}
@@ -305,10 +308,7 @@ static void queue_cast(struct dlm_rsb *r, struct dlm_lkb *lkb, int rv)
rv = -EDEADLK;
}
- lkb->lkb_lksb->sb_status = rv;
- lkb->lkb_lksb->sb_flags = lkb->lkb_sbflags;
-
- dlm_add_ast(lkb, AST_COMP, lkb->lkb_grmode);
+ dlm_add_cb(lkb, DLM_CB_CAST, lkb->lkb_grmode, rv, lkb->lkb_sbflags);
}
static inline void queue_cast_overlap(struct dlm_rsb *r, struct dlm_lkb *lkb)
@@ -319,13 +319,10 @@ static inline void queue_cast_overlap(struct dlm_rsb *r, struct dlm_lkb *lkb)
static void queue_bast(struct dlm_rsb *r, struct dlm_lkb *lkb, int rqmode)
{
- lkb->lkb_time_bast = ktime_get();
-
if (is_master_copy(lkb)) {
- lkb->lkb_bastmode = rqmode; /* printed by debugfs */
send_bast(r, lkb, rqmode);
} else {
- dlm_add_ast(lkb, AST_BAST, rqmode);
+ dlm_add_cb(lkb, DLM_CB_BAST, rqmode, 0, 0);
}
}
@@ -333,13 +330,94 @@ static void queue_bast(struct dlm_rsb *r, struct dlm_lkb *lkb, int rqmode)
* Basic operations on rsb's and lkb's
*/
-static struct dlm_rsb *create_rsb(struct dlm_ls *ls, char *name, int len)
+/* This is only called to add a reference when the code already holds
+ a valid reference to the rsb, so there's no need for locking. */
+
+static inline void hold_rsb(struct dlm_rsb *r)
+{
+ kref_get(&r->res_ref);
+}
+
+void dlm_hold_rsb(struct dlm_rsb *r)
+{
+ hold_rsb(r);
+}
+
+/* When all references to the rsb are gone it's transferred to
+ the tossed list for later disposal. */
+
+static void put_rsb(struct dlm_rsb *r)
+{
+ struct dlm_ls *ls = r->res_ls;
+ uint32_t bucket = r->res_bucket;
+
+ spin_lock(&ls->ls_rsbtbl[bucket].lock);
+ kref_put(&r->res_ref, toss_rsb);
+ spin_unlock(&ls->ls_rsbtbl[bucket].lock);
+}
+
+void dlm_put_rsb(struct dlm_rsb *r)
+{
+ put_rsb(r);
+}
+
+static int pre_rsb_struct(struct dlm_ls *ls)
+{
+ struct dlm_rsb *r1, *r2;
+ int count = 0;
+
+ spin_lock(&ls->ls_new_rsb_spin);
+ if (ls->ls_new_rsb_count > dlm_config.ci_new_rsb_count / 2) {
+ spin_unlock(&ls->ls_new_rsb_spin);
+ return 0;
+ }
+ spin_unlock(&ls->ls_new_rsb_spin);
+
+ r1 = dlm_allocate_rsb(ls);
+ r2 = dlm_allocate_rsb(ls);
+
+ spin_lock(&ls->ls_new_rsb_spin);
+ if (r1) {
+ list_add(&r1->res_hashchain, &ls->ls_new_rsb);
+ ls->ls_new_rsb_count++;
+ }
+ if (r2) {
+ list_add(&r2->res_hashchain, &ls->ls_new_rsb);
+ ls->ls_new_rsb_count++;
+ }
+ count = ls->ls_new_rsb_count;
+ spin_unlock(&ls->ls_new_rsb_spin);
+
+ if (!count)
+ return -ENOMEM;
+ return 0;
+}
+
+/* If ls->ls_new_rsb is empty, return -EAGAIN, so the caller can
+ unlock any spinlocks, go back and call pre_rsb_struct again.
+ Otherwise, take an rsb off the list and return it. */
+
+static int get_rsb_struct(struct dlm_ls *ls, char *name, int len,
+ struct dlm_rsb **r_ret)
{
struct dlm_rsb *r;
+ int count;
+
+ spin_lock(&ls->ls_new_rsb_spin);
+ if (list_empty(&ls->ls_new_rsb)) {
+ count = ls->ls_new_rsb_count;
+ spin_unlock(&ls->ls_new_rsb_spin);
+ log_debug(ls, "find_rsb retry %d %d %s",
+ count, dlm_config.ci_new_rsb_count, name);
+ return -EAGAIN;
+ }
- r = dlm_allocate_rsb(ls, len);
- if (!r)
- return NULL;
+ r = list_first_entry(&ls->ls_new_rsb, struct dlm_rsb, res_hashchain);
+ list_del(&r->res_hashchain);
+ /* Convert the empty list_head to a NULL rb_node for tree usage: */
+ memset(&r->res_hashnode, 0, sizeof(struct rb_node));
+ ls->ls_new_rsb_count--;
+ spin_unlock(&ls->ls_new_rsb_spin);
r->res_ls = ls;
r->res_length = len;
@@ -353,161 +431,696 @@ static struct dlm_rsb *create_rsb(struct dlm_ls *ls, char *name, int len)
INIT_LIST_HEAD(&r->res_root_list);
INIT_LIST_HEAD(&r->res_recover_list);
- return r;
+ *r_ret = r;
+ return 0;
}
-static int search_rsb_list(struct list_head *head, char *name, int len,
- unsigned int flags, struct dlm_rsb **r_ret)
+static int rsb_cmp(struct dlm_rsb *r, const char *name, int nlen)
{
- struct dlm_rsb *r;
- int error = 0;
+ char maxname[DLM_RESNAME_MAXLEN];
+
+ memset(maxname, 0, DLM_RESNAME_MAXLEN);
+ memcpy(maxname, name, nlen);
+ return memcmp(r->res_name, maxname, DLM_RESNAME_MAXLEN);
+}
- list_for_each_entry(r, head, res_hashchain) {
- if (len == r->res_length && !memcmp(name, r->res_name, len))
+int dlm_search_rsb_tree(struct rb_root *tree, char *name, int len,
+ struct dlm_rsb **r_ret)
+{
+ struct rb_node *node = tree->rb_node;
+ struct dlm_rsb *r;
+ int rc;
+
+ while (node) {
+ r = rb_entry(node, struct dlm_rsb, res_hashnode);
+ rc = rsb_cmp(r, name, len);
+ if (rc < 0)
+ node = node->rb_left;
+ else if (rc > 0)
+ node = node->rb_right;
+ else
goto found;
}
*r_ret = NULL;
return -EBADR;
found:
- if (r->res_nodeid && (flags & R_MASTER))
- error = -ENOTBLK;
*r_ret = r;
- return error;
+ return 0;
}
-static int _search_rsb(struct dlm_ls *ls, char *name, int len, int b,
- unsigned int flags, struct dlm_rsb **r_ret)
+static int rsb_insert(struct dlm_rsb *rsb, struct rb_root *tree)
{
- struct dlm_rsb *r;
+ struct rb_node **newn = &tree->rb_node;
+ struct rb_node *parent = NULL;
+ int rc;
+
+ while (*newn) {
+ struct dlm_rsb *cur = rb_entry(*newn, struct dlm_rsb,
+ res_hashnode);
+
+ parent = *newn;
+ rc = rsb_cmp(cur, rsb->res_name, rsb->res_length);
+ if (rc < 0)
+ newn = &parent->rb_left;
+ else if (rc > 0)
+ newn = &parent->rb_right;
+ else {
+ log_print("rsb_insert match");
+ dlm_dump_rsb(rsb);
+ dlm_dump_rsb(cur);
+ return -EEXIST;
+ }
+ }
+
+ rb_link_node(&rsb->res_hashnode, parent, newn);
+ rb_insert_color(&rsb->res_hashnode, tree);
+ return 0;
+}
+
+/*
+ * Find rsb in rsbtbl and potentially create/add one
+ *
+ * Delaying the release of rsb's has a similar benefit to applications keeping
+ * NL locks on an rsb, but without the guarantee that the cached master value
+ * will still be valid when the rsb is reused. Apps aren't always smart enough
+ * to keep NL locks on an rsb that they may lock again shortly; this can lead
+ * to excessive master lookups and removals if we don't delay the release.
+ *
+ * Searching for an rsb means looking through both the normal list and toss
+ * list. When found on the toss list the rsb is moved to the normal list with
+ * ref count of 1; when found on normal list the ref count is incremented.
+ *
+ * rsb's on the keep list are being used locally and refcounted.
+ * rsb's on the toss list are not being used locally, and are not refcounted.
+ *
+ * The toss list rsb's were either
+ * - previously used locally but not any more (were on keep list, then
+ * moved to toss list when last refcount dropped)
+ * - created and put on toss list as a directory record for a lookup
+ * (we are the dir node for the res, but are not using the res right now,
+ * but some other node is)
+ *
+ * The purpose of find_rsb() is to return a refcounted rsb for local use.
+ * So, if the given rsb is on the toss list, it is moved to the keep list
+ * before being returned.
+ *
+ * toss_rsb() happens when all local usage of the rsb is done, i.e. no
+ * more refcounts exist, so the rsb is moved from the keep list to the
+ * toss list.
+ *
+ * rsb's on both keep and toss lists are used for doing a name to master
+ * lookups. rsb's that are in use locally (and being refcounted) are on
+ * the keep list, rsb's that are not in use locally (not refcounted) and
+ * only exist for name/master lookups are on the toss list.
+ *
+ * rsb's on the toss list who's dir_nodeid is not local can have stale
+ * name/master mappings. So, remote requests on such rsb's can potentially
+ * return with an error, which means the mapping is stale and needs to
+ * be updated with a new lookup. (The idea behind MASTER UNCERTAIN and
+ * first_lkid is to keep only a single outstanding request on an rsb
+ * while that rsb has a potentially stale master.)
+ */
+
+static int find_rsb_dir(struct dlm_ls *ls, char *name, int len,
+ uint32_t hash, uint32_t b,
+ int dir_nodeid, int from_nodeid,
+ unsigned int flags, struct dlm_rsb **r_ret)
+{
+ struct dlm_rsb *r = NULL;
+ int our_nodeid = dlm_our_nodeid();
+ int from_local = 0;
+ int from_other = 0;
+ int from_dir = 0;
+ int create = 0;
int error;
- error = search_rsb_list(&ls->ls_rsbtbl[b].list, name, len, flags, &r);
- if (!error) {
- kref_get(&r->res_ref);
- goto out;
+ if (flags & R_RECEIVE_REQUEST) {
+ if (from_nodeid == dir_nodeid)
+ from_dir = 1;
+ else
+ from_other = 1;
+ } else if (flags & R_REQUEST) {
+ from_local = 1;
+ }
+
+ /*
+ * flags & R_RECEIVE_RECOVER is from dlm_recover_master_copy, so
+ * from_nodeid has sent us a lock in dlm_recover_locks, believing
+ * we're the new master. Our local recovery may not have set
+ * res_master_nodeid to our_nodeid yet, so allow either. Don't
+ * create the rsb; dlm_recover_process_copy() will handle EBADR
+ * by resending.
+ *
+ * If someone sends us a request, we are the dir node, and we do
+ * not find the rsb anywhere, then recreate it. This happens if
+ * someone sends us a request after we have removed/freed an rsb
+ * from our toss list. (They sent a request instead of lookup
+ * because they are using an rsb from their toss list.)
+ */
+
+ if (from_local || from_dir ||
+ (from_other && (dir_nodeid == our_nodeid))) {
+ create = 1;
+ }
+
+ retry:
+ if (create) {
+ error = pre_rsb_struct(ls);
+ if (error < 0)
+ goto out;
}
- error = search_rsb_list(&ls->ls_rsbtbl[b].toss, name, len, flags, &r);
+
+ spin_lock(&ls->ls_rsbtbl[b].lock);
+
+ error = dlm_search_rsb_tree(&ls->ls_rsbtbl[b].keep, name, len, &r);
if (error)
- goto out;
+ goto do_toss;
+
+ /*
+ * rsb is active, so we can't check master_nodeid without lock_rsb.
+ */
- list_move(&r->res_hashchain, &ls->ls_rsbtbl[b].list);
+ kref_get(&r->res_ref);
+ error = 0;
+ goto out_unlock;
- if (dlm_no_directory(ls))
- goto out;
- if (r->res_nodeid == -1) {
+ do_toss:
+ error = dlm_search_rsb_tree(&ls->ls_rsbtbl[b].toss, name, len, &r);
+ if (error)
+ goto do_new;
+
+ /*
+ * rsb found inactive (master_nodeid may be out of date unless
+ * we are the dir_nodeid or were the master) No other thread
+ * is using this rsb because it's on the toss list, so we can
+ * look at or update res_master_nodeid without lock_rsb.
+ */
+
+ if ((r->res_master_nodeid != our_nodeid) && from_other) {
+ /* our rsb was not master, and another node (not the dir node)
+ has sent us a request */
+ log_debug(ls, "find_rsb toss from_other %d master %d dir %d %s",
+ from_nodeid, r->res_master_nodeid, dir_nodeid,
+ r->res_name);
+ error = -ENOTBLK;
+ goto out_unlock;
+ }
+
+ if ((r->res_master_nodeid != our_nodeid) && from_dir) {
+ /* don't think this should ever happen */
+ log_error(ls, "find_rsb toss from_dir %d master %d",
+ from_nodeid, r->res_master_nodeid);
+ dlm_print_rsb(r);
+ /* fix it and go on */
+ r->res_master_nodeid = our_nodeid;
+ r->res_nodeid = 0;
rsb_clear_flag(r, RSB_MASTER_UNCERTAIN);
r->res_first_lkid = 0;
- } else if (r->res_nodeid > 0) {
+ }
+
+ if (from_local && (r->res_master_nodeid != our_nodeid)) {
+ /* Because we have held no locks on this rsb,
+ res_master_nodeid could have become stale. */
rsb_set_flag(r, RSB_MASTER_UNCERTAIN);
r->res_first_lkid = 0;
+ }
+
+ rb_erase(&r->res_hashnode, &ls->ls_rsbtbl[b].toss);
+ error = rsb_insert(r, &ls->ls_rsbtbl[b].keep);
+ goto out_unlock;
+
+
+ do_new:
+ /*
+ * rsb not found
+ */
+
+ if (error == -EBADR && !create)
+ goto out_unlock;
+
+ error = get_rsb_struct(ls, name, len, &r);
+ if (error == -EAGAIN) {
+ spin_unlock(&ls->ls_rsbtbl[b].lock);
+ goto retry;
+ }
+ if (error)
+ goto out_unlock;
+
+ r->res_hash = hash;
+ r->res_bucket = b;
+ r->res_dir_nodeid = dir_nodeid;
+ kref_init(&r->res_ref);
+
+ if (from_dir) {
+ /* want to see how often this happens */
+ log_debug(ls, "find_rsb new from_dir %d recreate %s",
+ from_nodeid, r->res_name);
+ r->res_master_nodeid = our_nodeid;
+ r->res_nodeid = 0;
+ goto out_add;
+ }
+
+ if (from_other && (dir_nodeid != our_nodeid)) {
+ /* should never happen */
+ log_error(ls, "find_rsb new from_other %d dir %d our %d %s",
+ from_nodeid, dir_nodeid, our_nodeid, r->res_name);
+ dlm_free_rsb(r);
+ r = NULL;
+ error = -ENOTBLK;
+ goto out_unlock;
+ }
+
+ if (from_other) {
+ log_debug(ls, "find_rsb new from_other %d dir %d %s",
+ from_nodeid, dir_nodeid, r->res_name);
+ }
+
+ if (dir_nodeid == our_nodeid) {
+ /* When we are the dir nodeid, we can set the master
+ node immediately */
+ r->res_master_nodeid = our_nodeid;
+ r->res_nodeid = 0;
} else {
- DLM_ASSERT(r->res_nodeid == 0, dlm_print_rsb(r););
- DLM_ASSERT(!rsb_flag(r, RSB_MASTER_UNCERTAIN),);
+ /* set_master will send_lookup to dir_nodeid */
+ r->res_master_nodeid = 0;
+ r->res_nodeid = -1;
}
+
+ out_add:
+ error = rsb_insert(r, &ls->ls_rsbtbl[b].keep);
+ out_unlock:
+ spin_unlock(&ls->ls_rsbtbl[b].lock);
out:
*r_ret = r;
return error;
}
-static int search_rsb(struct dlm_ls *ls, char *name, int len, int b,
- unsigned int flags, struct dlm_rsb **r_ret)
+/* During recovery, other nodes can send us new MSTCPY locks (from
+ dlm_recover_locks) before we've made ourself master (in
+ dlm_recover_masters). */
+
+static int find_rsb_nodir(struct dlm_ls *ls, char *name, int len,
+ uint32_t hash, uint32_t b,
+ int dir_nodeid, int from_nodeid,
+ unsigned int flags, struct dlm_rsb **r_ret)
{
+ struct dlm_rsb *r = NULL;
+ int our_nodeid = dlm_our_nodeid();
+ int recover = (flags & R_RECEIVE_RECOVER);
int error;
+
+ retry:
+ error = pre_rsb_struct(ls);
+ if (error < 0)
+ goto out;
+
spin_lock(&ls->ls_rsbtbl[b].lock);
- error = _search_rsb(ls, name, len, b, flags, r_ret);
+
+ error = dlm_search_rsb_tree(&ls->ls_rsbtbl[b].keep, name, len, &r);
+ if (error)
+ goto do_toss;
+
+ /*
+ * rsb is active, so we can't check master_nodeid without lock_rsb.
+ */
+
+ kref_get(&r->res_ref);
+ goto out_unlock;
+
+
+ do_toss:
+ error = dlm_search_rsb_tree(&ls->ls_rsbtbl[b].toss, name, len, &r);
+ if (error)
+ goto do_new;
+
+ /*
+ * rsb found inactive. No other thread is using this rsb because
+ * it's on the toss list, so we can look at or update
+ * res_master_nodeid without lock_rsb.
+ */
+
+ if (!recover && (r->res_master_nodeid != our_nodeid) && from_nodeid) {
+ /* our rsb is not master, and another node has sent us a
+ request; this should never happen */
+ log_error(ls, "find_rsb toss from_nodeid %d master %d dir %d",
+ from_nodeid, r->res_master_nodeid, dir_nodeid);
+ dlm_print_rsb(r);
+ error = -ENOTBLK;
+ goto out_unlock;
+ }
+
+ if (!recover && (r->res_master_nodeid != our_nodeid) &&
+ (dir_nodeid == our_nodeid)) {
+ /* our rsb is not master, and we are dir; may as well fix it;
+ this should never happen */
+ log_error(ls, "find_rsb toss our %d master %d dir %d",
+ our_nodeid, r->res_master_nodeid, dir_nodeid);
+ dlm_print_rsb(r);
+ r->res_master_nodeid = our_nodeid;
+ r->res_nodeid = 0;
+ }
+
+ rb_erase(&r->res_hashnode, &ls->ls_rsbtbl[b].toss);
+ error = rsb_insert(r, &ls->ls_rsbtbl[b].keep);
+ goto out_unlock;
+
+
+ do_new:
+ /*
+ * rsb not found
+ */
+
+ error = get_rsb_struct(ls, name, len, &r);
+ if (error == -EAGAIN) {
+ spin_unlock(&ls->ls_rsbtbl[b].lock);
+ goto retry;
+ }
+ if (error)
+ goto out_unlock;
+
+ r->res_hash = hash;
+ r->res_bucket = b;
+ r->res_dir_nodeid = dir_nodeid;
+ r->res_master_nodeid = dir_nodeid;
+ r->res_nodeid = (dir_nodeid == our_nodeid) ? 0 : dir_nodeid;
+ kref_init(&r->res_ref);
+
+ error = rsb_insert(r, &ls->ls_rsbtbl[b].keep);
+ out_unlock:
spin_unlock(&ls->ls_rsbtbl[b].lock);
+ out:
+ *r_ret = r;
return error;
}
+static int find_rsb(struct dlm_ls *ls, char *name, int len, int from_nodeid,
+ unsigned int flags, struct dlm_rsb **r_ret)
+{
+ uint32_t hash, b;
+ int dir_nodeid;
+
+ if (len > DLM_RESNAME_MAXLEN)
+ return -EINVAL;
+
+ hash = jhash(name, len, 0);
+ b = hash & (ls->ls_rsbtbl_size - 1);
+
+ dir_nodeid = dlm_hash2nodeid(ls, hash);
+
+ if (dlm_no_directory(ls))
+ return find_rsb_nodir(ls, name, len, hash, b, dir_nodeid,
+ from_nodeid, flags, r_ret);
+ else
+ return find_rsb_dir(ls, name, len, hash, b, dir_nodeid,
+ from_nodeid, flags, r_ret);
+}
+
+/* we have received a request and found that res_master_nodeid != our_nodeid,
+ so we need to return an error or make ourself the master */
+
+static int validate_master_nodeid(struct dlm_ls *ls, struct dlm_rsb *r,
+ int from_nodeid)
+{
+ if (dlm_no_directory(ls)) {
+ log_error(ls, "find_rsb keep from_nodeid %d master %d dir %d",
+ from_nodeid, r->res_master_nodeid,
+ r->res_dir_nodeid);
+ dlm_print_rsb(r);
+ return -ENOTBLK;
+ }
+
+ if (from_nodeid != r->res_dir_nodeid) {
+ /* our rsb is not master, and another node (not the dir node)
+ has sent us a request. this is much more common when our
+ master_nodeid is zero, so limit debug to non-zero. */
+
+ if (r->res_master_nodeid) {
+ log_debug(ls, "validate master from_other %d master %d "
+ "dir %d first %x %s", from_nodeid,
+ r->res_master_nodeid, r->res_dir_nodeid,
+ r->res_first_lkid, r->res_name);
+ }
+ return -ENOTBLK;
+ } else {
+ /* our rsb is not master, but the dir nodeid has sent us a
+ request; this could happen with master 0 / res_nodeid -1 */
+
+ if (r->res_master_nodeid) {
+ log_error(ls, "validate master from_dir %d master %d "
+ "first %x %s",
+ from_nodeid, r->res_master_nodeid,
+ r->res_first_lkid, r->res_name);
+ }
+
+ r->res_master_nodeid = dlm_our_nodeid();
+ r->res_nodeid = 0;
+ return 0;
+ }
+}
+
/*
- * Find rsb in rsbtbl and potentially create/add one
+ * We're the dir node for this res and another node wants to know the
+ * master nodeid. During normal operation (non recovery) this is only
+ * called from receive_lookup(); master lookups when the local node is
+ * the dir node are done by find_rsb().
*
- * Delaying the release of rsb's has a similar benefit to applications keeping
- * NL locks on an rsb, but without the guarantee that the cached master value
- * will still be valid when the rsb is reused. Apps aren't always smart enough
- * to keep NL locks on an rsb that they may lock again shortly; this can lead
- * to excessive master lookups and removals if we don't delay the release.
+ * normal operation, we are the dir node for a resource
+ * . _request_lock
+ * . set_master
+ * . send_lookup
+ * . receive_lookup
+ * . dlm_master_lookup flags 0
*
- * Searching for an rsb means looking through both the normal list and toss
- * list. When found on the toss list the rsb is moved to the normal list with
- * ref count of 1; when found on normal list the ref count is incremented.
+ * recover directory, we are rebuilding dir for all resources
+ * . dlm_recover_directory
+ * . dlm_rcom_names
+ * remote node sends back the rsb names it is master of and we are dir of
+ * . dlm_master_lookup RECOVER_DIR (fix_master 0, from_master 1)
+ * we either create new rsb setting remote node as master, or find existing
+ * rsb and set master to be the remote node.
+ *
+ * recover masters, we are finding the new master for resources
+ * . dlm_recover_masters
+ * . recover_master
+ * . dlm_send_rcom_lookup
+ * . receive_rcom_lookup
+ * . dlm_master_lookup RECOVER_MASTER (fix_master 1, from_master 0)
*/
-static int find_rsb(struct dlm_ls *ls, char *name, int namelen,
- unsigned int flags, struct dlm_rsb **r_ret)
+int dlm_master_lookup(struct dlm_ls *ls, int from_nodeid, char *name, int len,
+ unsigned int flags, int *r_nodeid, int *result)
{
- struct dlm_rsb *r = NULL, *tmp;
- uint32_t hash, bucket;
- int error = -EINVAL;
+ struct dlm_rsb *r = NULL;
+ uint32_t hash, b;
+ int from_master = (flags & DLM_LU_RECOVER_DIR);
+ int fix_master = (flags & DLM_LU_RECOVER_MASTER);
+ int our_nodeid = dlm_our_nodeid();
+ int dir_nodeid, error, toss_list = 0;
- if (namelen > DLM_RESNAME_MAXLEN)
- goto out;
+ if (len > DLM_RESNAME_MAXLEN)
+ return -EINVAL;
- if (dlm_no_directory(ls))
- flags |= R_CREATE;
+ if (from_nodeid == our_nodeid) {
+ log_error(ls, "dlm_master_lookup from our_nodeid %d flags %x",
+ our_nodeid, flags);
+ return -EINVAL;
+ }
- error = 0;
- hash = jhash(name, namelen, 0);
- bucket = hash & (ls->ls_rsbtbl_size - 1);
+ hash = jhash(name, len, 0);
+ b = hash & (ls->ls_rsbtbl_size - 1);
- error = search_rsb(ls, name, namelen, bucket, flags, &r);
- if (!error)
- goto out;
+ dir_nodeid = dlm_hash2nodeid(ls, hash);
+ if (dir_nodeid != our_nodeid) {
+ log_error(ls, "dlm_master_lookup from %d dir %d our %d h %x %d",
+ from_nodeid, dir_nodeid, our_nodeid, hash,
+ ls->ls_num_nodes);
+ *r_nodeid = -1;
+ return -EINVAL;
+ }
- if (error == -EBADR && !(flags & R_CREATE))
- goto out;
+ retry:
+ error = pre_rsb_struct(ls);
+ if (error < 0)
+ return error;
- /* the rsb was found but wasn't a master copy */
- if (error == -ENOTBLK)
- goto out;
+ spin_lock(&ls->ls_rsbtbl[b].lock);
+ error = dlm_search_rsb_tree(&ls->ls_rsbtbl[b].keep, name, len, &r);
+ if (!error) {
+ /* because the rsb is active, we need to lock_rsb before
+ checking/changing re_master_nodeid */
- error = -ENOMEM;
- r = create_rsb(ls, name, namelen);
- if (!r)
- goto out;
+ hold_rsb(r);
+ spin_unlock(&ls->ls_rsbtbl[b].lock);
+ lock_rsb(r);
+ goto found;
+ }
- r->res_hash = hash;
- r->res_bucket = bucket;
- r->res_nodeid = -1;
- kref_init(&r->res_ref);
+ error = dlm_search_rsb_tree(&ls->ls_rsbtbl[b].toss, name, len, &r);
+ if (error)
+ goto not_found;
- /* With no directory, the master can be set immediately */
- if (dlm_no_directory(ls)) {
- int nodeid = dlm_dir_nodeid(r);
- if (nodeid == dlm_our_nodeid())
- nodeid = 0;
- r->res_nodeid = nodeid;
+ /* because the rsb is inactive (on toss list), it's not refcounted
+ and lock_rsb is not used, but is protected by the rsbtbl lock */
+
+ toss_list = 1;
+ found:
+ if (r->res_dir_nodeid != our_nodeid) {
+ /* should not happen, but may as well fix it and carry on */
+ log_error(ls, "dlm_master_lookup res_dir %d our %d %s",
+ r->res_dir_nodeid, our_nodeid, r->res_name);
+ r->res_dir_nodeid = our_nodeid;
+ }
+
+ if (fix_master && dlm_is_removed(ls, r->res_master_nodeid)) {
+ /* Recovery uses this function to set a new master when
+ the previous master failed. Setting NEW_MASTER will
+ force dlm_recover_masters to call recover_master on this
+ rsb even though the res_nodeid is no longer removed. */
+
+ r->res_master_nodeid = from_nodeid;
+ r->res_nodeid = from_nodeid;
+ rsb_set_flag(r, RSB_NEW_MASTER);
+
+ if (toss_list) {
+ /* I don't think we should ever find it on toss list. */
+ log_error(ls, "dlm_master_lookup fix_master on toss");
+ dlm_dump_rsb(r);
+ }
}
- spin_lock(&ls->ls_rsbtbl[bucket].lock);
- error = _search_rsb(ls, name, namelen, bucket, 0, &tmp);
- if (!error) {
- spin_unlock(&ls->ls_rsbtbl[bucket].lock);
+ if (from_master && (r->res_master_nodeid != from_nodeid)) {
+ /* this will happen if from_nodeid became master during
+ a previous recovery cycle, and we aborted the previous
+ cycle before recovering this master value */
+
+ log_limit(ls, "dlm_master_lookup from_master %d "
+ "master_nodeid %d res_nodeid %d first %x %s",
+ from_nodeid, r->res_master_nodeid, r->res_nodeid,
+ r->res_first_lkid, r->res_name);
+
+ if (r->res_master_nodeid == our_nodeid) {
+ log_error(ls, "from_master %d our_master", from_nodeid);
+ dlm_dump_rsb(r);
+ dlm_send_rcom_lookup_dump(r, from_nodeid);
+ goto out_found;
+ }
+
+ r->res_master_nodeid = from_nodeid;
+ r->res_nodeid = from_nodeid;
+ rsb_set_flag(r, RSB_NEW_MASTER);
+ }
+
+ if (!r->res_master_nodeid) {
+ /* this will happen if recovery happens while we're looking
+ up the master for this rsb */
+
+ log_debug(ls, "dlm_master_lookup master 0 to %d first %x %s",
+ from_nodeid, r->res_first_lkid, r->res_name);
+ r->res_master_nodeid = from_nodeid;
+ r->res_nodeid = from_nodeid;
+ }
+
+ if (!from_master && !fix_master &&
+ (r->res_master_nodeid == from_nodeid)) {
+ /* this can happen when the master sends remove, the dir node
+ finds the rsb on the keep list and ignores the remove,
+ and the former master sends a lookup */
+
+ log_limit(ls, "dlm_master_lookup from master %d flags %x "
+ "first %x %s", from_nodeid, flags,
+ r->res_first_lkid, r->res_name);
+ }
+
+ out_found:
+ *r_nodeid = r->res_master_nodeid;
+ if (result)
+ *result = DLM_LU_MATCH;
+
+ if (toss_list) {
+ r->res_toss_time = jiffies;
+ /* the rsb was inactive (on toss list) */
+ spin_unlock(&ls->ls_rsbtbl[b].lock);
+ } else {
+ /* the rsb was active */
+ unlock_rsb(r);
+ put_rsb(r);
+ }
+ return 0;
+
+ not_found:
+ error = get_rsb_struct(ls, name, len, &r);
+ if (error == -EAGAIN) {
+ spin_unlock(&ls->ls_rsbtbl[b].lock);
+ goto retry;
+ }
+ if (error)
+ goto out_unlock;
+
+ r->res_hash = hash;
+ r->res_bucket = b;
+ r->res_dir_nodeid = our_nodeid;
+ r->res_master_nodeid = from_nodeid;
+ r->res_nodeid = from_nodeid;
+ kref_init(&r->res_ref);
+ r->res_toss_time = jiffies;
+
+ error = rsb_insert(r, &ls->ls_rsbtbl[b].toss);
+ if (error) {
+ /* should never happen */
dlm_free_rsb(r);
- r = tmp;
- goto out;
+ spin_unlock(&ls->ls_rsbtbl[b].lock);
+ goto retry;
}
- list_add(&r->res_hashchain, &ls->ls_rsbtbl[bucket].list);
- spin_unlock(&ls->ls_rsbtbl[bucket].lock);
+
+ if (result)
+ *result = DLM_LU_ADD;
+ *r_nodeid = from_nodeid;
error = 0;
- out:
- *r_ret = r;
+ out_unlock:
+ spin_unlock(&ls->ls_rsbtbl[b].lock);
return error;
}
-/* This is only called to add a reference when the code already holds
- a valid reference to the rsb, so there's no need for locking. */
-
-static inline void hold_rsb(struct dlm_rsb *r)
+static void dlm_dump_rsb_hash(struct dlm_ls *ls, uint32_t hash)
{
- kref_get(&r->res_ref);
+ struct rb_node *n;
+ struct dlm_rsb *r;
+ int i;
+
+ for (i = 0; i < ls->ls_rsbtbl_size; i++) {
+ spin_lock(&ls->ls_rsbtbl[i].lock);
+ for (n = rb_first(&ls->ls_rsbtbl[i].keep); n; n = rb_next(n)) {
+ r = rb_entry(n, struct dlm_rsb, res_hashnode);
+ if (r->res_hash == hash)
+ dlm_dump_rsb(r);
+ }
+ spin_unlock(&ls->ls_rsbtbl[i].lock);
+ }
}
-void dlm_hold_rsb(struct dlm_rsb *r)
+void dlm_dump_rsb_name(struct dlm_ls *ls, char *name, int len)
{
- hold_rsb(r);
+ struct dlm_rsb *r = NULL;
+ uint32_t hash, b;
+ int error;
+
+ hash = jhash(name, len, 0);
+ b = hash & (ls->ls_rsbtbl_size - 1);
+
+ spin_lock(&ls->ls_rsbtbl[b].lock);
+ error = dlm_search_rsb_tree(&ls->ls_rsbtbl[b].keep, name, len, &r);
+ if (!error)
+ goto out_dump;
+
+ error = dlm_search_rsb_tree(&ls->ls_rsbtbl[b].toss, name, len, &r);
+ if (error)
+ goto out;
+ out_dump:
+ dlm_dump_rsb(r);
+ out:
+ spin_unlock(&ls->ls_rsbtbl[b].lock);
}
static void toss_rsb(struct kref *kref)
@@ -517,32 +1130,16 @@ static void toss_rsb(struct kref *kref)
DLM_ASSERT(list_empty(&r->res_root_list), dlm_print_rsb(r););
kref_init(&r->res_ref);
- list_move(&r->res_hashchain, &ls->ls_rsbtbl[r->res_bucket].toss);
+ rb_erase(&r->res_hashnode, &ls->ls_rsbtbl[r->res_bucket].keep);
+ rsb_insert(r, &ls->ls_rsbtbl[r->res_bucket].toss);
r->res_toss_time = jiffies;
+ ls->ls_rsbtbl[r->res_bucket].flags |= DLM_RTF_SHRINK;
if (r->res_lvbptr) {
dlm_free_lvb(r->res_lvbptr);
r->res_lvbptr = NULL;
}
}
-/* When all references to the rsb are gone it's transfered to
- the tossed list for later disposal. */
-
-static void put_rsb(struct dlm_rsb *r)
-{
- struct dlm_ls *ls = r->res_ls;
- uint32_t bucket = r->res_bucket;
-
- spin_lock(&ls->ls_rsbtbl[bucket].lock);
- kref_put(&r->res_ref, toss_rsb);
- spin_unlock(&ls->ls_rsbtbl[bucket].lock);
-}
-
-void dlm_put_rsb(struct dlm_rsb *r)
-{
- put_rsb(r);
-}
-
/* See comment for unhold_lkb */
static void unhold_rsb(struct dlm_rsb *r)
@@ -586,9 +1183,8 @@ static void detach_lkb(struct dlm_lkb *lkb)
static int create_lkb(struct dlm_ls *ls, struct dlm_lkb **lkb_ret)
{
- struct dlm_lkb *lkb, *tmp;
- uint32_t lkid = 0;
- uint16_t bucket;
+ struct dlm_lkb *lkb;
+ int rv;
lkb = dlm_allocate_lkb(ls);
if (!lkb)
@@ -600,59 +1196,36 @@ static int create_lkb(struct dlm_ls *ls, struct dlm_lkb **lkb_ret)
INIT_LIST_HEAD(&lkb->lkb_ownqueue);
INIT_LIST_HEAD(&lkb->lkb_rsb_lookup);
INIT_LIST_HEAD(&lkb->lkb_time_list);
+ INIT_LIST_HEAD(&lkb->lkb_cb_list);
+ mutex_init(&lkb->lkb_cb_mutex);
+ INIT_WORK(&lkb->lkb_cb_work, dlm_callback_work);
- get_random_bytes(&bucket, sizeof(bucket));
- bucket &= (ls->ls_lkbtbl_size - 1);
+ idr_preload(GFP_NOFS);
+ spin_lock(&ls->ls_lkbidr_spin);
+ rv = idr_alloc(&ls->ls_lkbidr, lkb, 1, 0, GFP_NOWAIT);
+ if (rv >= 0)
+ lkb->lkb_id = rv;
+ spin_unlock(&ls->ls_lkbidr_spin);
+ idr_preload_end();
- write_lock(&ls->ls_lkbtbl[bucket].lock);
-
- /* counter can roll over so we must verify lkid is not in use */
-
- while (lkid == 0) {
- lkid = (bucket << 16) | ls->ls_lkbtbl[bucket].counter++;
-
- list_for_each_entry(tmp, &ls->ls_lkbtbl[bucket].list,
- lkb_idtbl_list) {
- if (tmp->lkb_id != lkid)
- continue;
- lkid = 0;
- break;
- }
+ if (rv < 0) {
+ log_error(ls, "create_lkb idr error %d", rv);
+ return rv;
}
- lkb->lkb_id = lkid;
- list_add(&lkb->lkb_idtbl_list, &ls->ls_lkbtbl[bucket].list);
- write_unlock(&ls->ls_lkbtbl[bucket].lock);
-
*lkb_ret = lkb;
return 0;
}
-static struct dlm_lkb *__find_lkb(struct dlm_ls *ls, uint32_t lkid)
-{
- struct dlm_lkb *lkb;
- uint16_t bucket = (lkid >> 16);
-
- list_for_each_entry(lkb, &ls->ls_lkbtbl[bucket].list, lkb_idtbl_list) {
- if (lkb->lkb_id == lkid)
- return lkb;
- }
- return NULL;
-}
-
static int find_lkb(struct dlm_ls *ls, uint32_t lkid, struct dlm_lkb **lkb_ret)
{
struct dlm_lkb *lkb;
- uint16_t bucket = (lkid >> 16);
-
- if (bucket >= ls->ls_lkbtbl_size)
- return -EBADSLT;
- read_lock(&ls->ls_lkbtbl[bucket].lock);
- lkb = __find_lkb(ls, lkid);
+ spin_lock(&ls->ls_lkbidr_spin);
+ lkb = idr_find(&ls->ls_lkbidr, lkid);
if (lkb)
kref_get(&lkb->lkb_ref);
- read_unlock(&ls->ls_lkbtbl[bucket].lock);
+ spin_unlock(&ls->ls_lkbidr_spin);
*lkb_ret = lkb;
return lkb ? 0 : -ENOENT;
@@ -673,12 +1246,12 @@ static void kill_lkb(struct kref *kref)
static int __put_lkb(struct dlm_ls *ls, struct dlm_lkb *lkb)
{
- uint16_t bucket = (lkb->lkb_id >> 16);
+ uint32_t lkid = lkb->lkb_id;
- write_lock(&ls->ls_lkbtbl[bucket].lock);
+ spin_lock(&ls->ls_lkbidr_spin);
if (kref_put(&lkb->lkb_ref, kill_lkb)) {
- list_del(&lkb->lkb_idtbl_list);
- write_unlock(&ls->ls_lkbtbl[bucket].lock);
+ idr_remove(&ls->ls_lkbidr, lkid);
+ spin_unlock(&ls->ls_lkbidr_spin);
detach_lkb(lkb);
@@ -688,7 +1261,7 @@ static int __put_lkb(struct dlm_ls *ls, struct dlm_lkb *lkb)
dlm_free_lkb(lkb);
return 1;
} else {
- write_unlock(&ls->ls_lkbtbl[bucket].lock);
+ spin_unlock(&ls->ls_lkbidr_spin);
return 0;
}
}
@@ -804,10 +1377,80 @@ static int msg_reply_type(int mstype)
return -1;
}
+static int nodeid_warned(int nodeid, int num_nodes, int *warned)
+{
+ int i;
+
+ for (i = 0; i < num_nodes; i++) {
+ if (!warned[i]) {
+ warned[i] = nodeid;
+ return 0;
+ }
+ if (warned[i] == nodeid)
+ return 1;
+ }
+ return 0;
+}
+
+void dlm_scan_waiters(struct dlm_ls *ls)
+{
+ struct dlm_lkb *lkb;
+ ktime_t zero = ktime_set(0, 0);
+ s64 us;
+ s64 debug_maxus = 0;
+ u32 debug_scanned = 0;
+ u32 debug_expired = 0;
+ int num_nodes = 0;
+ int *warned = NULL;
+
+ if (!dlm_config.ci_waitwarn_us)
+ return;
+
+ mutex_lock(&ls->ls_waiters_mutex);
+
+ list_for_each_entry(lkb, &ls->ls_waiters, lkb_wait_reply) {
+ if (ktime_equal(lkb->lkb_wait_time, zero))
+ continue;
+
+ debug_scanned++;
+
+ us = ktime_to_us(ktime_sub(ktime_get(), lkb->lkb_wait_time));
+
+ if (us < dlm_config.ci_waitwarn_us)
+ continue;
+
+ lkb->lkb_wait_time = zero;
+
+ debug_expired++;
+ if (us > debug_maxus)
+ debug_maxus = us;
+
+ if (!num_nodes) {
+ num_nodes = ls->ls_num_nodes;
+ warned = kzalloc(num_nodes * sizeof(int), GFP_KERNEL);
+ }
+ if (!warned)
+ continue;
+ if (nodeid_warned(lkb->lkb_wait_nodeid, num_nodes, warned))
+ continue;
+
+ log_error(ls, "waitwarn %x %lld %d us check connection to "
+ "node %d", lkb->lkb_id, (long long)us,
+ dlm_config.ci_waitwarn_us, lkb->lkb_wait_nodeid);
+ }
+ mutex_unlock(&ls->ls_waiters_mutex);
+ kfree(warned);
+
+ if (debug_expired)
+ log_debug(ls, "scan_waiters %u warn %u over %d us max %lld us",
+ debug_scanned, debug_expired,
+ dlm_config.ci_waitwarn_us, (long long)debug_maxus);
+}
+
/* add/remove lkb from global waiters list of lkb's waiting for
a reply from a remote node */
-static int add_to_waiters(struct dlm_lkb *lkb, int mstype)
+static int add_to_waiters(struct dlm_lkb *lkb, int mstype, int to_nodeid)
{
struct dlm_ls *ls = lkb->lkb_resource->res_ls;
int error = 0;
@@ -847,6 +1490,8 @@ static int add_to_waiters(struct dlm_lkb *lkb, int mstype)
lkb->lkb_wait_count++;
lkb->lkb_wait_type = mstype;
+ lkb->lkb_wait_time = ktime_get();
+ lkb->lkb_wait_nodeid = to_nodeid; /* for debugging */
hold_lkb(lkb);
list_add(&lkb->lkb_wait_reply, &ls->ls_waiters);
out:
@@ -920,8 +1565,9 @@ static int _remove_from_waiters(struct dlm_lkb *lkb, int mstype,
goto out_del;
}
- log_error(ls, "remwait error %x reply %d flags %x no wait_type",
- lkb->lkb_id, mstype, lkb->lkb_flags);
+ log_error(ls, "remwait error %x remote %d %x msg %d flags %x no wait",
+ lkb->lkb_id, ms ? ms->m_header.h_nodeid : 0, lkb->lkb_remid,
+ mstype, lkb->lkb_flags);
return -1;
out_del:
@@ -966,69 +1612,192 @@ static int remove_from_waiters_ms(struct dlm_lkb *lkb, struct dlm_message *ms)
struct dlm_ls *ls = lkb->lkb_resource->res_ls;
int error;
- if (ms != &ls->ls_stub_ms)
+ if (ms->m_flags != DLM_IFL_STUB_MS)
mutex_lock(&ls->ls_waiters_mutex);
error = _remove_from_waiters(lkb, ms->m_type, ms);
- if (ms != &ls->ls_stub_ms)
+ if (ms->m_flags != DLM_IFL_STUB_MS)
mutex_unlock(&ls->ls_waiters_mutex);
return error;
}
-static void dir_remove(struct dlm_rsb *r)
+/* If there's an rsb for the same resource being removed, ensure
+ that the remove message is sent before the new lookup message.
+ It should be rare to need a delay here, but if not, then it may
+ be worthwhile to add a proper wait mechanism rather than a delay. */
+
+static void wait_pending_remove(struct dlm_rsb *r)
{
- int to_nodeid;
+ struct dlm_ls *ls = r->res_ls;
+ restart:
+ spin_lock(&ls->ls_remove_spin);
+ if (ls->ls_remove_len &&
+ !rsb_cmp(r, ls->ls_remove_name, ls->ls_remove_len)) {
+ log_debug(ls, "delay lookup for remove dir %d %s",
+ r->res_dir_nodeid, r->res_name);
+ spin_unlock(&ls->ls_remove_spin);
+ msleep(1);
+ goto restart;
+ }
+ spin_unlock(&ls->ls_remove_spin);
+}
- if (dlm_no_directory(r->res_ls))
+/*
+ * ls_remove_spin protects ls_remove_name and ls_remove_len which are
+ * read by other threads in wait_pending_remove. ls_remove_names
+ * and ls_remove_lens are only used by the scan thread, so they do
+ * not need protection.
+ */
+
+static void shrink_bucket(struct dlm_ls *ls, int b)
+{
+ struct rb_node *n, *next;
+ struct dlm_rsb *r;
+ char *name;
+ int our_nodeid = dlm_our_nodeid();
+ int remote_count = 0;
+ int need_shrink = 0;
+ int i, len, rv;
+
+ memset(&ls->ls_remove_lens, 0, sizeof(int) * DLM_REMOVE_NAMES_MAX);
+
+ spin_lock(&ls->ls_rsbtbl[b].lock);
+
+ if (!(ls->ls_rsbtbl[b].flags & DLM_RTF_SHRINK)) {
+ spin_unlock(&ls->ls_rsbtbl[b].lock);
return;
+ }
- to_nodeid = dlm_dir_nodeid(r);
- if (to_nodeid != dlm_our_nodeid())
- send_remove(r);
+ for (n = rb_first(&ls->ls_rsbtbl[b].toss); n; n = next) {
+ next = rb_next(n);
+ r = rb_entry(n, struct dlm_rsb, res_hashnode);
+
+ /* If we're the directory record for this rsb, and
+ we're not the master of it, then we need to wait
+ for the master node to send us a dir remove for
+ before removing the dir record. */
+
+ if (!dlm_no_directory(ls) &&
+ (r->res_master_nodeid != our_nodeid) &&
+ (dlm_dir_nodeid(r) == our_nodeid)) {
+ continue;
+ }
+
+ need_shrink = 1;
+
+ if (!time_after_eq(jiffies, r->res_toss_time +
+ dlm_config.ci_toss_secs * HZ)) {
+ continue;
+ }
+
+ if (!dlm_no_directory(ls) &&
+ (r->res_master_nodeid == our_nodeid) &&
+ (dlm_dir_nodeid(r) != our_nodeid)) {
+
+ /* We're the master of this rsb but we're not
+ the directory record, so we need to tell the
+ dir node to remove the dir record. */
+
+ ls->ls_remove_lens[remote_count] = r->res_length;
+ memcpy(ls->ls_remove_names[remote_count], r->res_name,
+ DLM_RESNAME_MAXLEN);
+ remote_count++;
+
+ if (remote_count >= DLM_REMOVE_NAMES_MAX)
+ break;
+ continue;
+ }
+
+ if (!kref_put(&r->res_ref, kill_rsb)) {
+ log_error(ls, "tossed rsb in use %s", r->res_name);
+ continue;
+ }
+
+ rb_erase(&r->res_hashnode, &ls->ls_rsbtbl[b].toss);
+ dlm_free_rsb(r);
+ }
+
+ if (need_shrink)
+ ls->ls_rsbtbl[b].flags |= DLM_RTF_SHRINK;
else
- dlm_dir_remove_entry(r->res_ls, to_nodeid,
- r->res_name, r->res_length);
-}
+ ls->ls_rsbtbl[b].flags &= ~DLM_RTF_SHRINK;
+ spin_unlock(&ls->ls_rsbtbl[b].lock);
-/* FIXME: shouldn't this be able to exit as soon as one non-due rsb is
- found since they are in order of newest to oldest? */
+ /*
+ * While searching for rsb's to free, we found some that require
+ * remote removal. We leave them in place and find them again here
+ * so there is a very small gap between removing them from the toss
+ * list and sending the removal. Keeping this gap small is
+ * important to keep us (the master node) from being out of sync
+ * with the remote dir node for very long.
+ *
+ * From the time the rsb is removed from toss until just after
+ * send_remove, the rsb name is saved in ls_remove_name. A new
+ * lookup checks this to ensure that a new lookup message for the
+ * same resource name is not sent just before the remove message.
+ */
-static int shrink_bucket(struct dlm_ls *ls, int b)
-{
- struct dlm_rsb *r;
- int count = 0, found;
+ for (i = 0; i < remote_count; i++) {
+ name = ls->ls_remove_names[i];
+ len = ls->ls_remove_lens[i];
- for (;;) {
- found = 0;
spin_lock(&ls->ls_rsbtbl[b].lock);
- list_for_each_entry_reverse(r, &ls->ls_rsbtbl[b].toss,
- res_hashchain) {
- if (!time_after_eq(jiffies, r->res_toss_time +
- dlm_config.ci_toss_secs * HZ))
- continue;
- found = 1;
- break;
+ rv = dlm_search_rsb_tree(&ls->ls_rsbtbl[b].toss, name, len, &r);
+ if (rv) {
+ spin_unlock(&ls->ls_rsbtbl[b].lock);
+ log_debug(ls, "remove_name not toss %s", name);
+ continue;
}
- if (!found) {
+ if (r->res_master_nodeid != our_nodeid) {
spin_unlock(&ls->ls_rsbtbl[b].lock);
- break;
+ log_debug(ls, "remove_name master %d dir %d our %d %s",
+ r->res_master_nodeid, r->res_dir_nodeid,
+ our_nodeid, name);
+ continue;
}
- if (kref_put(&r->res_ref, kill_rsb)) {
- list_del(&r->res_hashchain);
+ if (r->res_dir_nodeid == our_nodeid) {
+ /* should never happen */
spin_unlock(&ls->ls_rsbtbl[b].lock);
+ log_error(ls, "remove_name dir %d master %d our %d %s",
+ r->res_dir_nodeid, r->res_master_nodeid,
+ our_nodeid, name);
+ continue;
+ }
- if (is_master(r))
- dir_remove(r);
- dlm_free_rsb(r);
- count++;
- } else {
+ if (!time_after_eq(jiffies, r->res_toss_time +
+ dlm_config.ci_toss_secs * HZ)) {
spin_unlock(&ls->ls_rsbtbl[b].lock);
- log_error(ls, "tossed rsb in use %s", r->res_name);
+ log_debug(ls, "remove_name toss_time %lu now %lu %s",
+ r->res_toss_time, jiffies, name);
+ continue;
}
- }
- return count;
+ if (!kref_put(&r->res_ref, kill_rsb)) {
+ spin_unlock(&ls->ls_rsbtbl[b].lock);
+ log_error(ls, "remove_name in use %s", name);
+ continue;
+ }
+
+ rb_erase(&r->res_hashnode, &ls->ls_rsbtbl[b].toss);
+
+ /* block lookup of same name until we've sent remove */
+ spin_lock(&ls->ls_remove_spin);
+ ls->ls_remove_len = len;
+ memcpy(ls->ls_remove_name, name, DLM_RESNAME_MAXLEN);
+ spin_unlock(&ls->ls_remove_spin);
+ spin_unlock(&ls->ls_rsbtbl[b].lock);
+
+ send_remove(r);
+
+ /* allow lookup of name again */
+ spin_lock(&ls->ls_remove_spin);
+ ls->ls_remove_len = 0;
+ memset(ls->ls_remove_name, 0, DLM_RESNAME_MAXLEN);
+ spin_unlock(&ls->ls_remove_spin);
+
+ dlm_free_rsb(r);
+ }
}
void dlm_scan_rsbs(struct dlm_ls *ls)
@@ -1162,6 +1931,16 @@ void dlm_adjust_timeouts(struct dlm_ls *ls)
list_for_each_entry(lkb, &ls->ls_timeout, lkb_time_list)
lkb->lkb_timestamp = ktime_add_us(lkb->lkb_timestamp, adj_us);
mutex_unlock(&ls->ls_timeout_mutex);
+
+ if (!dlm_config.ci_waitwarn_us)
+ return;
+
+ mutex_lock(&ls->ls_waiters_mutex);
+ list_for_each_entry(lkb, &ls->ls_waiters, lkb_wait_reply) {
+ if (ktime_to_us(lkb->lkb_wait_time))
+ lkb->lkb_wait_time = ktime_get();
+ }
+ mutex_unlock(&ls->ls_waiters_mutex);
}
/* lkb is master or local copy */
@@ -1260,8 +2039,8 @@ static void set_lvb_lock_pc(struct dlm_rsb *r, struct dlm_lkb *lkb,
b = dlm_lvb_operations[lkb->lkb_grmode + 1][lkb->lkb_rqmode + 1];
if (b == 1) {
int len = receive_extralen(ms);
- if (len > DLM_RESNAME_MAXLEN)
- len = DLM_RESNAME_MAXLEN;
+ if (len > r->res_ls->ls_lvblen)
+ len = r->res_ls->ls_lvblen;
memcpy(lkb->lkb_lvbptr, ms->m_extra, len);
lkb->lkb_lvbseq = ms->m_lvbseq;
}
@@ -1344,13 +2123,13 @@ static void _grant_lock(struct dlm_rsb *r, struct dlm_lkb *lkb)
}
lkb->lkb_rqmode = DLM_LOCK_IV;
+ lkb->lkb_highbast = 0;
}
static void grant_lock(struct dlm_rsb *r, struct dlm_lkb *lkb)
{
set_lvb_lock(r, lkb);
_grant_lock(r, lkb);
- lkb->lkb_highbast = 0;
}
static void grant_lock_pc(struct dlm_rsb *r, struct dlm_lkb *lkb,
@@ -1381,14 +2160,8 @@ static void grant_lock_pending(struct dlm_rsb *r, struct dlm_lkb *lkb)
ALTPR/ALTCW: our rqmode may have been changed to PR or CW to become
compatible with other granted locks */
-static void munge_demoted(struct dlm_lkb *lkb, struct dlm_message *ms)
+static void munge_demoted(struct dlm_lkb *lkb)
{
- if (ms->m_type != DLM_MSG_CONVERT_REPLY) {
- log_print("munge_demoted %x invalid reply type %d",
- lkb->lkb_id, ms->m_type);
- return;
- }
-
if (lkb->lkb_rqmode == DLM_LOCK_IV || lkb->lkb_grmode == DLM_LOCK_IV) {
log_print("munge_demoted %x invalid modes gr %d rq %d",
lkb->lkb_id, lkb->lkb_grmode, lkb->lkb_rqmode);
@@ -1516,10 +2289,14 @@ static int conversion_deadlock_detect(struct dlm_rsb *r, struct dlm_lkb *lkb2)
* immediate request, it is 0 if called later, after the lock has been
* queued.
*
+ * recover is 1 if dlm_recover_grant() is trying to grant conversions
+ * after recovery.
+ *
* References are from chapter 6 of "VAXcluster Principles" by Roy Davis
*/
-static int _can_be_granted(struct dlm_rsb *r, struct dlm_lkb *lkb, int now)
+static int _can_be_granted(struct dlm_rsb *r, struct dlm_lkb *lkb, int now,
+ int recover)
{
int8_t conv = (lkb->lkb_grmode != DLM_LOCK_IV);
@@ -1551,7 +2328,7 @@ static int _can_be_granted(struct dlm_rsb *r, struct dlm_lkb *lkb, int now)
*/
if (queue_conflict(&r->res_grantqueue, lkb))
- goto out;
+ return 0;
/*
* 6-3: By default, a conversion request is immediately granted if the
@@ -1560,7 +2337,24 @@ static int _can_be_granted(struct dlm_rsb *r, struct dlm_lkb *lkb, int now)
*/
if (queue_conflict(&r->res_convertqueue, lkb))
- goto out;
+ return 0;
+
+ /*
+ * The RECOVER_GRANT flag means dlm_recover_grant() is granting
+ * locks for a recovered rsb, on which lkb's have been rebuilt.
+ * The lkb's may have been rebuilt on the queues in a different
+ * order than they were in on the previous master. So, granting
+ * queued conversions in order after recovery doesn't make sense
+ * since the order hasn't been preserved anyway. The new order
+ * could also have created a new "in place" conversion deadlock.
+ * (e.g. old, failed master held granted EX, with PR->EX, NL->EX.
+ * After recovery, there would be no granted locks, and possibly
+ * NL->EX, PR->EX, an in-place conversion deadlock.) So, after
+ * recovery, grant conversions without considering order.
+ */
+
+ if (conv && recover)
+ return 1;
/*
* 6-5: But the default algorithm for deciding whether to grant or
@@ -1589,6 +2383,18 @@ static int _can_be_granted(struct dlm_rsb *r, struct dlm_lkb *lkb, int now)
return 1;
/*
+ * Even if the convert is compat with all granted locks,
+ * QUECVT forces it behind other locks on the convert queue.
+ */
+
+ if (now && conv && (lkb->lkb_exflags & DLM_LKF_QUECVT)) {
+ if (list_empty(&r->res_convertqueue))
+ return 1;
+ else
+ return 0;
+ }
+
+ /*
* The NOORDER flag is set to avoid the standard vms rules on grant
* order.
*/
@@ -1631,12 +2437,12 @@ static int _can_be_granted(struct dlm_rsb *r, struct dlm_lkb *lkb, int now)
if (!now && !conv && list_empty(&r->res_convertqueue) &&
first_in_list(lkb, &r->res_waitqueue))
return 1;
- out:
+
return 0;
}
static int can_be_granted(struct dlm_rsb *r, struct dlm_lkb *lkb, int now,
- int *err)
+ int recover, int *err)
{
int rv;
int8_t alt = 0, rqmode = lkb->lkb_rqmode;
@@ -1645,7 +2451,7 @@ static int can_be_granted(struct dlm_rsb *r, struct dlm_lkb *lkb, int now,
if (err)
*err = 0;
- rv = _can_be_granted(r, lkb, now);
+ rv = _can_be_granted(r, lkb, now, recover);
if (rv)
goto out;
@@ -1686,7 +2492,7 @@ static int can_be_granted(struct dlm_rsb *r, struct dlm_lkb *lkb, int now,
if (alt) {
lkb->lkb_rqmode = alt;
- rv = _can_be_granted(r, lkb, now);
+ rv = _can_be_granted(r, lkb, now, 0);
if (rv)
lkb->lkb_sbflags |= DLM_SBF_ALTMODE;
else
@@ -1706,9 +2512,11 @@ static int can_be_granted(struct dlm_rsb *r, struct dlm_lkb *lkb, int now,
/* Returns the highest requested mode of all blocked conversions; sets
cw if there's a blocked conversion to DLM_LOCK_CW. */
-static int grant_pending_convert(struct dlm_rsb *r, int high, int *cw)
+static int grant_pending_convert(struct dlm_rsb *r, int high, int *cw,
+ unsigned int *count)
{
struct dlm_lkb *lkb, *s;
+ int recover = rsb_flag(r, RSB_RECOVER_GRANT);
int hi, demoted, quit, grant_restart, demote_restart;
int deadlk;
@@ -1722,9 +2530,11 @@ static int grant_pending_convert(struct dlm_rsb *r, int high, int *cw)
demoted = is_demoted(lkb);
deadlk = 0;
- if (can_be_granted(r, lkb, 0, &deadlk)) {
+ if (can_be_granted(r, lkb, 0, recover, &deadlk)) {
grant_lock_pending(r, lkb);
grant_restart = 1;
+ if (count)
+ (*count)++;
continue;
}
@@ -1758,14 +2568,17 @@ static int grant_pending_convert(struct dlm_rsb *r, int high, int *cw)
return max_t(int, high, hi);
}
-static int grant_pending_wait(struct dlm_rsb *r, int high, int *cw)
+static int grant_pending_wait(struct dlm_rsb *r, int high, int *cw,
+ unsigned int *count)
{
struct dlm_lkb *lkb, *s;
list_for_each_entry_safe(lkb, s, &r->res_waitqueue, lkb_statequeue) {
- if (can_be_granted(r, lkb, 0, NULL))
+ if (can_be_granted(r, lkb, 0, 0, NULL)) {
grant_lock_pending(r, lkb);
- else {
+ if (count)
+ (*count)++;
+ } else {
high = max_t(int, lkb->lkb_rqmode, high);
if (lkb->lkb_rqmode == DLM_LOCK_CW)
*cw = 1;
@@ -1794,16 +2607,20 @@ static int lock_requires_bast(struct dlm_lkb *gr, int high, int cw)
return 0;
}
-static void grant_pending_locks(struct dlm_rsb *r)
+static void grant_pending_locks(struct dlm_rsb *r, unsigned int *count)
{
struct dlm_lkb *lkb, *s;
int high = DLM_LOCK_IV;
int cw = 0;
- DLM_ASSERT(is_master(r), dlm_dump_rsb(r););
+ if (!is_master(r)) {
+ log_print("grant_pending_locks r nodeid %d", r->res_nodeid);
+ dlm_dump_rsb(r);
+ return;
+ }
- high = grant_pending_convert(r, high, &cw);
- high = grant_pending_wait(r, high, &cw);
+ high = grant_pending_convert(r, high, &cw, count);
+ high = grant_pending_wait(r, high, &cw, count);
if (high == DLM_LOCK_IV)
return;
@@ -1888,8 +2705,7 @@ static void send_blocking_asts_all(struct dlm_rsb *r, struct dlm_lkb *lkb)
static int set_master(struct dlm_rsb *r, struct dlm_lkb *lkb)
{
- struct dlm_ls *ls = r->res_ls;
- int i, error, dir_nodeid, ret_nodeid, our_nodeid = dlm_our_nodeid();
+ int our_nodeid = dlm_our_nodeid();
if (rsb_flag(r, RSB_MASTER_UNCERTAIN)) {
rsb_clear_flag(r, RSB_MASTER_UNCERTAIN);
@@ -1903,53 +2719,37 @@ static int set_master(struct dlm_rsb *r, struct dlm_lkb *lkb)
return 1;
}
- if (r->res_nodeid == 0) {
+ if (r->res_master_nodeid == our_nodeid) {
lkb->lkb_nodeid = 0;
return 0;
}
- if (r->res_nodeid > 0) {
- lkb->lkb_nodeid = r->res_nodeid;
+ if (r->res_master_nodeid) {
+ lkb->lkb_nodeid = r->res_master_nodeid;
return 0;
}
- DLM_ASSERT(r->res_nodeid == -1, dlm_dump_rsb(r););
-
- dir_nodeid = dlm_dir_nodeid(r);
-
- if (dir_nodeid != our_nodeid) {
- r->res_first_lkid = lkb->lkb_id;
- send_lookup(r, lkb);
- return 1;
- }
-
- for (i = 0; i < 2; i++) {
- /* It's possible for dlm_scand to remove an old rsb for
- this same resource from the toss list, us to create
- a new one, look up the master locally, and find it
- already exists just before dlm_scand does the
- dir_remove() on the previous rsb. */
-
- error = dlm_dir_lookup(ls, our_nodeid, r->res_name,
- r->res_length, &ret_nodeid);
- if (!error)
- break;
- log_debug(ls, "dir_lookup error %d %s", error, r->res_name);
- schedule();
- }
- if (error && error != -EEXIST)
- return error;
-
- if (ret_nodeid == our_nodeid) {
- r->res_first_lkid = 0;
+ if (dlm_dir_nodeid(r) == our_nodeid) {
+ /* This is a somewhat unusual case; find_rsb will usually
+ have set res_master_nodeid when dir nodeid is local, but
+ there are cases where we become the dir node after we've
+ past find_rsb and go through _request_lock again.
+ confirm_master() or process_lookup_list() needs to be
+ called after this. */
+ log_debug(r->res_ls, "set_master %x self master %d dir %d %s",
+ lkb->lkb_id, r->res_master_nodeid, r->res_dir_nodeid,
+ r->res_name);
+ r->res_master_nodeid = our_nodeid;
r->res_nodeid = 0;
lkb->lkb_nodeid = 0;
- } else {
- r->res_first_lkid = lkb->lkb_id;
- r->res_nodeid = ret_nodeid;
- lkb->lkb_nodeid = ret_nodeid;
+ return 0;
}
- return 0;
+
+ wait_pending_remove(r);
+
+ r->res_first_lkid = lkb->lkb_id;
+ send_lookup(r, lkb);
+ return 1;
}
static void process_lookup_list(struct dlm_rsb *r)
@@ -2274,7 +3074,7 @@ static int do_request(struct dlm_rsb *r, struct dlm_lkb *lkb)
{
int error = 0;
- if (can_be_granted(r, lkb, 1, NULL)) {
+ if (can_be_granted(r, lkb, 1, 0, NULL)) {
grant_lock(r, lkb);
queue_cast(r, lkb, 0);
goto out;
@@ -2314,7 +3114,7 @@ static int do_convert(struct dlm_rsb *r, struct dlm_lkb *lkb)
/* changing an existing lock may allow others to be granted */
- if (can_be_granted(r, lkb, 1, &deadlk)) {
+ if (can_be_granted(r, lkb, 1, 0, &deadlk)) {
grant_lock(r, lkb);
queue_cast(r, lkb, 0);
goto out;
@@ -2326,9 +3126,6 @@ static int do_convert(struct dlm_rsb *r, struct dlm_lkb *lkb)
if (deadlk) {
/* it's left on the granted queue */
- log_debug(r->res_ls, "deadlock %x node %d sts%d g%d r%d %s",
- lkb->lkb_id, lkb->lkb_nodeid, lkb->lkb_status,
- lkb->lkb_grmode, lkb->lkb_rqmode, r->res_name);
revert_lock(r, lkb);
queue_cast(r, lkb, -EDEADLK);
error = -EDEADLK;
@@ -2342,8 +3139,8 @@ static int do_convert(struct dlm_rsb *r, struct dlm_lkb *lkb)
before we try again to grant this one. */
if (is_demoted(lkb)) {
- grant_pending_convert(r, DLM_LOCK_IV, NULL);
- if (_can_be_granted(r, lkb, 1)) {
+ grant_pending_convert(r, DLM_LOCK_IV, NULL, NULL);
+ if (_can_be_granted(r, lkb, 1, 0)) {
grant_lock(r, lkb);
queue_cast(r, lkb, 0);
goto out;
@@ -2370,7 +3167,7 @@ static void do_convert_effects(struct dlm_rsb *r, struct dlm_lkb *lkb,
{
switch (error) {
case 0:
- grant_pending_locks(r);
+ grant_pending_locks(r, NULL);
/* grant_pending_locks also sends basts */
break;
case -EAGAIN:
@@ -2393,11 +3190,11 @@ static int do_unlock(struct dlm_rsb *r, struct dlm_lkb *lkb)
static void do_unlock_effects(struct dlm_rsb *r, struct dlm_lkb *lkb,
int error)
{
- grant_pending_locks(r);
+ grant_pending_locks(r, NULL);
}
/* returns: 0 did nothing, -DLM_ECANCEL canceled lock */
-
+
static int do_cancel(struct dlm_rsb *r, struct dlm_lkb *lkb)
{
int error;
@@ -2414,7 +3211,7 @@ static void do_cancel_effects(struct dlm_rsb *r, struct dlm_lkb *lkb,
int error)
{
if (error)
- grant_pending_locks(r);
+ grant_pending_locks(r, NULL);
}
/*
@@ -2521,11 +3318,11 @@ static int request_lock(struct dlm_ls *ls, struct dlm_lkb *lkb, char *name,
error = validate_lock_args(ls, lkb, args);
if (error)
- goto out;
+ return error;
- error = find_rsb(ls, name, len, R_CREATE, &r);
+ error = find_rsb(ls, name, len, 0, R_REQUEST, &r);
if (error)
- goto out;
+ return error;
lock_rsb(r);
@@ -2536,8 +3333,6 @@ static int request_lock(struct dlm_ls *ls, struct dlm_lkb *lkb, char *name,
unlock_rsb(r);
put_rsb(r);
-
- out:
return error;
}
@@ -2819,9 +3614,9 @@ static void send_args(struct dlm_rsb *r, struct dlm_lkb *lkb,
not from lkb fields */
if (lkb->lkb_bastfn)
- ms->m_asts |= AST_BAST;
+ ms->m_asts |= DLM_CB_BAST;
if (lkb->lkb_astfn)
- ms->m_asts |= AST_COMP;
+ ms->m_asts |= DLM_CB_CAST;
/* compare with switch in create_message; send_remove() doesn't
use send_args() */
@@ -2849,12 +3644,12 @@ static int send_common(struct dlm_rsb *r, struct dlm_lkb *lkb, int mstype)
struct dlm_mhandle *mh;
int to_nodeid, error;
- error = add_to_waiters(lkb, mstype);
+ to_nodeid = r->res_nodeid;
+
+ error = add_to_waiters(lkb, mstype, to_nodeid);
if (error)
return error;
- to_nodeid = r->res_nodeid;
-
error = create_message(r, lkb, to_nodeid, mstype, &ms, &mh);
if (error)
goto fail;
@@ -2885,9 +3680,9 @@ static int send_convert(struct dlm_rsb *r, struct dlm_lkb *lkb)
/* down conversions go without a reply from the master */
if (!error && down_conversion(lkb)) {
remove_from_waiters(lkb, DLM_MSG_CONVERT_REPLY);
+ r->res_ls->ls_stub_ms.m_flags = DLM_IFL_STUB_MS;
r->res_ls->ls_stub_ms.m_type = DLM_MSG_CONVERT_REPLY;
r->res_ls->ls_stub_ms.m_result = 0;
- r->res_ls->ls_stub_ms.m_flags = lkb->lkb_flags;
__receive_convert_reply(r, lkb, &r->res_ls->ls_stub_ms);
}
@@ -2956,12 +3751,12 @@ static int send_lookup(struct dlm_rsb *r, struct dlm_lkb *lkb)
struct dlm_mhandle *mh;
int to_nodeid, error;
- error = add_to_waiters(lkb, DLM_MSG_LOOKUP);
+ to_nodeid = dlm_dir_nodeid(r);
+
+ error = add_to_waiters(lkb, DLM_MSG_LOOKUP, to_nodeid);
if (error)
return error;
- to_nodeid = dlm_dir_nodeid(r);
-
error = create_message(r, NULL, to_nodeid, DLM_MSG_LOOKUP, &ms, &mh);
if (error)
goto fail;
@@ -3075,6 +3870,9 @@ static void receive_flags(struct dlm_lkb *lkb, struct dlm_message *ms)
static void receive_flags_reply(struct dlm_lkb *lkb, struct dlm_message *ms)
{
+ if (ms->m_flags == DLM_IFL_STUB_MS)
+ return;
+
lkb->lkb_sbflags = ms->m_sbflags;
lkb->lkb_flags = (lkb->lkb_flags & 0xFFFF0000) |
(ms->m_flags & 0x0000FFFF);
@@ -3096,8 +3894,8 @@ static int receive_lvb(struct dlm_ls *ls, struct dlm_lkb *lkb,
if (!lkb->lkb_lvbptr)
return -ENOMEM;
len = receive_extralen(ms);
- if (len > DLM_RESNAME_MAXLEN)
- len = DLM_RESNAME_MAXLEN;
+ if (len > ls->ls_lvblen)
+ len = ls->ls_lvblen;
memcpy(lkb->lkb_lvbptr, ms->m_extra, len);
}
return 0;
@@ -3122,8 +3920,8 @@ static int receive_request_args(struct dlm_ls *ls, struct dlm_lkb *lkb,
lkb->lkb_grmode = DLM_LOCK_IV;
lkb->lkb_rqmode = ms->m_rqmode;
- lkb->lkb_bastfn = (ms->m_asts & AST_BAST) ? &fake_bastfn : NULL;
- lkb->lkb_astfn = (ms->m_asts & AST_COMP) ? &fake_astfn : NULL;
+ lkb->lkb_bastfn = (ms->m_asts & DLM_CB_BAST) ? &fake_bastfn : NULL;
+ lkb->lkb_astfn = (ms->m_asts & DLM_CB_CAST) ? &fake_astfn : NULL;
if (lkb->lkb_exflags & DLM_LKF_VALBLK) {
/* lkb was just created so there won't be an lvb yet */
@@ -3212,11 +4010,72 @@ static int validate_message(struct dlm_lkb *lkb, struct dlm_message *ms)
return error;
}
-static void receive_request(struct dlm_ls *ls, struct dlm_message *ms)
+static void send_repeat_remove(struct dlm_ls *ls, char *ms_name, int len)
+{
+ char name[DLM_RESNAME_MAXLEN + 1];
+ struct dlm_message *ms;
+ struct dlm_mhandle *mh;
+ struct dlm_rsb *r;
+ uint32_t hash, b;
+ int rv, dir_nodeid;
+
+ memset(name, 0, sizeof(name));
+ memcpy(name, ms_name, len);
+
+ hash = jhash(name, len, 0);
+ b = hash & (ls->ls_rsbtbl_size - 1);
+
+ dir_nodeid = dlm_hash2nodeid(ls, hash);
+
+ log_error(ls, "send_repeat_remove dir %d %s", dir_nodeid, name);
+
+ spin_lock(&ls->ls_rsbtbl[b].lock);
+ rv = dlm_search_rsb_tree(&ls->ls_rsbtbl[b].keep, name, len, &r);
+ if (!rv) {
+ spin_unlock(&ls->ls_rsbtbl[b].lock);
+ log_error(ls, "repeat_remove on keep %s", name);
+ return;
+ }
+
+ rv = dlm_search_rsb_tree(&ls->ls_rsbtbl[b].toss, name, len, &r);
+ if (!rv) {
+ spin_unlock(&ls->ls_rsbtbl[b].lock);
+ log_error(ls, "repeat_remove on toss %s", name);
+ return;
+ }
+
+ /* use ls->remove_name2 to avoid conflict with shrink? */
+
+ spin_lock(&ls->ls_remove_spin);
+ ls->ls_remove_len = len;
+ memcpy(ls->ls_remove_name, name, DLM_RESNAME_MAXLEN);
+ spin_unlock(&ls->ls_remove_spin);
+ spin_unlock(&ls->ls_rsbtbl[b].lock);
+
+ rv = _create_message(ls, sizeof(struct dlm_message) + len,
+ dir_nodeid, DLM_MSG_REMOVE, &ms, &mh);
+ if (rv)
+ return;
+
+ memcpy(ms->m_extra, name, len);
+ ms->m_hash = hash;
+
+ send_message(mh, ms);
+
+ spin_lock(&ls->ls_remove_spin);
+ ls->ls_remove_len = 0;
+ memset(ls->ls_remove_name, 0, DLM_RESNAME_MAXLEN);
+ spin_unlock(&ls->ls_remove_spin);
+}
+
+static int receive_request(struct dlm_ls *ls, struct dlm_message *ms)
{
struct dlm_lkb *lkb;
struct dlm_rsb *r;
- int error, namelen;
+ int from_nodeid;
+ int error, namelen = 0;
+
+ from_nodeid = ms->m_header.h_nodeid;
error = create_lkb(ls, &lkb);
if (error)
@@ -3230,9 +4089,16 @@ static void receive_request(struct dlm_ls *ls, struct dlm_message *ms)
goto fail;
}
+ /* The dir node is the authority on whether we are the master
+ for this rsb or not, so if the master sends us a request, we should
+ recreate the rsb if we've destroyed it. This race happens when we
+ send a remove message to the dir node at the same time that the dir
+ node sends us a request for the rsb. */
+
namelen = receive_extralen(ms);
- error = find_rsb(ls, ms->m_extra, namelen, R_MASTER, &r);
+ error = find_rsb(ls, ms->m_extra, namelen, from_nodeid,
+ R_RECEIVE_REQUEST, &r);
if (error) {
__put_lkb(ls, lkb);
goto fail;
@@ -3240,6 +4106,16 @@ static void receive_request(struct dlm_ls *ls, struct dlm_message *ms)
lock_rsb(r);
+ if (r->res_master_nodeid != dlm_our_nodeid()) {
+ error = validate_master_nodeid(ls, r, from_nodeid);
+ if (error) {
+ unlock_rsb(r);
+ put_rsb(r);
+ __put_lkb(ls, lkb);
+ goto fail;
+ }
+ }
+
attach_lkb(r, lkb);
error = do_request(r, lkb);
send_request_reply(r, lkb, error);
@@ -3252,14 +4128,40 @@ static void receive_request(struct dlm_ls *ls, struct dlm_message *ms)
error = 0;
if (error)
dlm_put_lkb(lkb);
- return;
+ return 0;
fail:
+ /* TODO: instead of returning ENOTBLK, add the lkb to res_lookup
+ and do this receive_request again from process_lookup_list once
+ we get the lookup reply. This would avoid a many repeated
+ ENOTBLK request failures when the lookup reply designating us
+ as master is delayed. */
+
+ /* We could repeatedly return -EBADR here if our send_remove() is
+ delayed in being sent/arriving/being processed on the dir node.
+ Another node would repeatedly lookup up the master, and the dir
+ node would continue returning our nodeid until our send_remove
+ took effect.
+
+ We send another remove message in case our previous send_remove
+ was lost/ignored/missed somehow. */
+
+ if (error != -ENOTBLK) {
+ log_limit(ls, "receive_request %x from %d %d",
+ ms->m_lkid, from_nodeid, error);
+ }
+
+ if (namelen && error == -EBADR) {
+ send_repeat_remove(ls, ms->m_extra, namelen);
+ msleep(1000);
+ }
+
setup_stub_lkb(ls, ms);
send_request_reply(&ls->ls_stub_rsb, &ls->ls_stub_lkb, error);
+ return error;
}
-static void receive_convert(struct dlm_ls *ls, struct dlm_message *ms)
+static int receive_convert(struct dlm_ls *ls, struct dlm_message *ms)
{
struct dlm_lkb *lkb;
struct dlm_rsb *r;
@@ -3269,6 +4171,15 @@ static void receive_convert(struct dlm_ls *ls, struct dlm_message *ms)
if (error)
goto fail;
+ if (lkb->lkb_remid != ms->m_lkid) {
+ log_error(ls, "receive_convert %x remid %x recover_seq %llu "
+ "remote %d %x", lkb->lkb_id, lkb->lkb_remid,
+ (unsigned long long)lkb->lkb_recover_seq,
+ ms->m_header.h_nodeid, ms->m_lkid);
+ error = -ENOENT;
+ goto fail;
+ }
+
r = lkb->lkb_resource;
hold_rsb(r);
@@ -3296,14 +4207,15 @@ static void receive_convert(struct dlm_ls *ls, struct dlm_message *ms)
unlock_rsb(r);
put_rsb(r);
dlm_put_lkb(lkb);
- return;
+ return 0;
fail:
setup_stub_lkb(ls, ms);
send_convert_reply(&ls->ls_stub_rsb, &ls->ls_stub_lkb, error);
+ return error;
}
-static void receive_unlock(struct dlm_ls *ls, struct dlm_message *ms)
+static int receive_unlock(struct dlm_ls *ls, struct dlm_message *ms)
{
struct dlm_lkb *lkb;
struct dlm_rsb *r;
@@ -3313,6 +4225,14 @@ static void receive_unlock(struct dlm_ls *ls, struct dlm_message *ms)
if (error)
goto fail;
+ if (lkb->lkb_remid != ms->m_lkid) {
+ log_error(ls, "receive_unlock %x remid %x remote %d %x",
+ lkb->lkb_id, lkb->lkb_remid,
+ ms->m_header.h_nodeid, ms->m_lkid);
+ error = -ENOENT;
+ goto fail;
+ }
+
r = lkb->lkb_resource;
hold_rsb(r);
@@ -3337,14 +4257,15 @@ static void receive_unlock(struct dlm_ls *ls, struct dlm_message *ms)
unlock_rsb(r);
put_rsb(r);
dlm_put_lkb(lkb);
- return;
+ return 0;
fail:
setup_stub_lkb(ls, ms);
send_unlock_reply(&ls->ls_stub_rsb, &ls->ls_stub_lkb, error);
+ return error;
}
-static void receive_cancel(struct dlm_ls *ls, struct dlm_message *ms)
+static int receive_cancel(struct dlm_ls *ls, struct dlm_message *ms)
{
struct dlm_lkb *lkb;
struct dlm_rsb *r;
@@ -3372,25 +4293,23 @@ static void receive_cancel(struct dlm_ls *ls, struct dlm_message *ms)
unlock_rsb(r);
put_rsb(r);
dlm_put_lkb(lkb);
- return;
+ return 0;
fail:
setup_stub_lkb(ls, ms);
send_cancel_reply(&ls->ls_stub_rsb, &ls->ls_stub_lkb, error);
+ return error;
}
-static void receive_grant(struct dlm_ls *ls, struct dlm_message *ms)
+static int receive_grant(struct dlm_ls *ls, struct dlm_message *ms)
{
struct dlm_lkb *lkb;
struct dlm_rsb *r;
int error;
error = find_lkb(ls, ms->m_remid, &lkb);
- if (error) {
- log_debug(ls, "receive_grant from %d no lkb %x",
- ms->m_header.h_nodeid, ms->m_remid);
- return;
- }
+ if (error)
+ return error;
r = lkb->lkb_resource;
@@ -3410,20 +4329,18 @@ static void receive_grant(struct dlm_ls *ls, struct dlm_message *ms)
unlock_rsb(r);
put_rsb(r);
dlm_put_lkb(lkb);
+ return 0;
}
-static void receive_bast(struct dlm_ls *ls, struct dlm_message *ms)
+static int receive_bast(struct dlm_ls *ls, struct dlm_message *ms)
{
struct dlm_lkb *lkb;
struct dlm_rsb *r;
int error;
error = find_lkb(ls, ms->m_remid, &lkb);
- if (error) {
- log_debug(ls, "receive_bast from %d no lkb %x",
- ms->m_header.h_nodeid, ms->m_remid);
- return;
- }
+ if (error)
+ return error;
r = lkb->lkb_resource;
@@ -3435,57 +4352,120 @@ static void receive_bast(struct dlm_ls *ls, struct dlm_message *ms)
goto out;
queue_bast(r, lkb, ms->m_bastmode);
+ lkb->lkb_highbast = ms->m_bastmode;
out:
unlock_rsb(r);
put_rsb(r);
dlm_put_lkb(lkb);
+ return 0;
}
static void receive_lookup(struct dlm_ls *ls, struct dlm_message *ms)
{
- int len, error, ret_nodeid, dir_nodeid, from_nodeid, our_nodeid;
+ int len, error, ret_nodeid, from_nodeid, our_nodeid;
from_nodeid = ms->m_header.h_nodeid;
our_nodeid = dlm_our_nodeid();
len = receive_extralen(ms);
- dir_nodeid = dlm_hash2nodeid(ls, ms->m_hash);
- if (dir_nodeid != our_nodeid) {
- log_error(ls, "lookup dir_nodeid %d from %d",
- dir_nodeid, from_nodeid);
- error = -EINVAL;
- ret_nodeid = -1;
- goto out;
- }
-
- error = dlm_dir_lookup(ls, from_nodeid, ms->m_extra, len, &ret_nodeid);
+ error = dlm_master_lookup(ls, from_nodeid, ms->m_extra, len, 0,
+ &ret_nodeid, NULL);
/* Optimization: we're master so treat lookup as a request */
if (!error && ret_nodeid == our_nodeid) {
receive_request(ls, ms);
return;
}
- out:
send_lookup_reply(ls, ms, ret_nodeid, error);
}
static void receive_remove(struct dlm_ls *ls, struct dlm_message *ms)
{
- int len, dir_nodeid, from_nodeid;
+ char name[DLM_RESNAME_MAXLEN+1];
+ struct dlm_rsb *r;
+ uint32_t hash, b;
+ int rv, len, dir_nodeid, from_nodeid;
from_nodeid = ms->m_header.h_nodeid;
len = receive_extralen(ms);
+ if (len > DLM_RESNAME_MAXLEN) {
+ log_error(ls, "receive_remove from %d bad len %d",
+ from_nodeid, len);
+ return;
+ }
+
dir_nodeid = dlm_hash2nodeid(ls, ms->m_hash);
if (dir_nodeid != dlm_our_nodeid()) {
- log_error(ls, "remove dir entry dir_nodeid %d from %d",
- dir_nodeid, from_nodeid);
+ log_error(ls, "receive_remove from %d bad nodeid %d",
+ from_nodeid, dir_nodeid);
return;
}
- dlm_dir_remove_entry(ls, from_nodeid, ms->m_extra, len);
+ /* Look for name on rsbtbl.toss, if it's there, kill it.
+ If it's on rsbtbl.keep, it's being used, and we should ignore this
+ message. This is an expected race between the dir node sending a
+ request to the master node at the same time as the master node sends
+ a remove to the dir node. The resolution to that race is for the
+ dir node to ignore the remove message, and the master node to
+ recreate the master rsb when it gets a request from the dir node for
+ an rsb it doesn't have. */
+
+ memset(name, 0, sizeof(name));
+ memcpy(name, ms->m_extra, len);
+
+ hash = jhash(name, len, 0);
+ b = hash & (ls->ls_rsbtbl_size - 1);
+
+ spin_lock(&ls->ls_rsbtbl[b].lock);
+
+ rv = dlm_search_rsb_tree(&ls->ls_rsbtbl[b].toss, name, len, &r);
+ if (rv) {
+ /* verify the rsb is on keep list per comment above */
+ rv = dlm_search_rsb_tree(&ls->ls_rsbtbl[b].keep, name, len, &r);
+ if (rv) {
+ /* should not happen */
+ log_error(ls, "receive_remove from %d not found %s",
+ from_nodeid, name);
+ spin_unlock(&ls->ls_rsbtbl[b].lock);
+ return;
+ }
+ if (r->res_master_nodeid != from_nodeid) {
+ /* should not happen */
+ log_error(ls, "receive_remove keep from %d master %d",
+ from_nodeid, r->res_master_nodeid);
+ dlm_print_rsb(r);
+ spin_unlock(&ls->ls_rsbtbl[b].lock);
+ return;
+ }
+
+ log_debug(ls, "receive_remove from %d master %d first %x %s",
+ from_nodeid, r->res_master_nodeid, r->res_first_lkid,
+ name);
+ spin_unlock(&ls->ls_rsbtbl[b].lock);
+ return;
+ }
+
+ if (r->res_master_nodeid != from_nodeid) {
+ log_error(ls, "receive_remove toss from %d master %d",
+ from_nodeid, r->res_master_nodeid);
+ dlm_print_rsb(r);
+ spin_unlock(&ls->ls_rsbtbl[b].lock);
+ return;
+ }
+
+ if (kref_put(&r->res_ref, kill_rsb)) {
+ rb_erase(&r->res_hashnode, &ls->ls_rsbtbl[b].toss);
+ spin_unlock(&ls->ls_rsbtbl[b].lock);
+ dlm_free_rsb(r);
+ } else {
+ log_error(ls, "receive_remove from %d rsb ref error",
+ from_nodeid);
+ dlm_print_rsb(r);
+ spin_unlock(&ls->ls_rsbtbl[b].lock);
+ }
}
static void receive_purge(struct dlm_ls *ls, struct dlm_message *ms)
@@ -3493,18 +4473,16 @@ static void receive_purge(struct dlm_ls *ls, struct dlm_message *ms)
do_purge(ls, ms->m_nodeid, ms->m_pid);
}
-static void receive_request_reply(struct dlm_ls *ls, struct dlm_message *ms)
+static int receive_request_reply(struct dlm_ls *ls, struct dlm_message *ms)
{
struct dlm_lkb *lkb;
struct dlm_rsb *r;
int error, mstype, result;
+ int from_nodeid = ms->m_header.h_nodeid;
error = find_lkb(ls, ms->m_remid, &lkb);
- if (error) {
- log_debug(ls, "receive_request_reply from %d no lkb %x",
- ms->m_header.h_nodeid, ms->m_remid);
- return;
- }
+ if (error)
+ return error;
r = lkb->lkb_resource;
hold_rsb(r);
@@ -3516,14 +4494,19 @@ static void receive_request_reply(struct dlm_ls *ls, struct dlm_message *ms)
mstype = lkb->lkb_wait_type;
error = remove_from_waiters(lkb, DLM_MSG_REQUEST_REPLY);
- if (error)
+ if (error) {
+ log_error(ls, "receive_request_reply %x remote %d %x result %d",
+ lkb->lkb_id, from_nodeid, ms->m_lkid, ms->m_result);
+ dlm_dump_rsb(r);
goto out;
+ }
/* Optimization: the dir node was also the master, so it took our
lookup as a request and sent request reply instead of lookup reply */
if (mstype == DLM_MSG_LOOKUP) {
- r->res_nodeid = ms->m_header.h_nodeid;
- lkb->lkb_nodeid = r->res_nodeid;
+ r->res_master_nodeid = from_nodeid;
+ r->res_nodeid = from_nodeid;
+ lkb->lkb_nodeid = from_nodeid;
}
/* this is the value returned from do_request() on the master */
@@ -3557,18 +4540,30 @@ static void receive_request_reply(struct dlm_ls *ls, struct dlm_message *ms)
case -EBADR:
case -ENOTBLK:
/* find_rsb failed to find rsb or rsb wasn't master */
- log_debug(ls, "receive_request_reply %x %x master diff %d %d",
- lkb->lkb_id, lkb->lkb_flags, r->res_nodeid, result);
- r->res_nodeid = -1;
- lkb->lkb_nodeid = -1;
+ log_limit(ls, "receive_request_reply %x from %d %d "
+ "master %d dir %d first %x %s", lkb->lkb_id,
+ from_nodeid, result, r->res_master_nodeid,
+ r->res_dir_nodeid, r->res_first_lkid, r->res_name);
+
+ if (r->res_dir_nodeid != dlm_our_nodeid() &&
+ r->res_master_nodeid != dlm_our_nodeid()) {
+ /* cause _request_lock->set_master->send_lookup */
+ r->res_master_nodeid = 0;
+ r->res_nodeid = -1;
+ lkb->lkb_nodeid = -1;
+ }
if (is_overlap(lkb)) {
/* we'll ignore error in cancel/unlock reply */
queue_cast_overlap(r, lkb);
confirm_master(r, result);
unhold_lkb(lkb); /* undoes create_lkb() */
- } else
+ } else {
_request_lock(r, lkb);
+
+ if (r->res_master_nodeid == dlm_our_nodeid())
+ confirm_master(r, 0);
+ }
break;
default:
@@ -3595,6 +4590,7 @@ static void receive_request_reply(struct dlm_ls *ls, struct dlm_message *ms)
unlock_rsb(r);
put_rsb(r);
dlm_put_lkb(lkb);
+ return 0;
}
static void __receive_convert_reply(struct dlm_rsb *r, struct dlm_lkb *lkb,
@@ -3617,7 +4613,7 @@ static void __receive_convert_reply(struct dlm_rsb *r, struct dlm_lkb *lkb,
/* convert was queued on remote master */
receive_flags_reply(lkb, ms);
if (is_demoted(lkb))
- munge_demoted(lkb, ms);
+ munge_demoted(lkb);
del_lkb(r, lkb);
add_lkb(r, lkb, DLM_LKSTS_CONVERT);
add_timeout(lkb);
@@ -3627,14 +4623,17 @@ static void __receive_convert_reply(struct dlm_rsb *r, struct dlm_lkb *lkb,
/* convert was granted on remote master */
receive_flags_reply(lkb, ms);
if (is_demoted(lkb))
- munge_demoted(lkb, ms);
+ munge_demoted(lkb);
grant_lock_pc(r, lkb, ms);
queue_cast(r, lkb, 0);
break;
default:
- log_error(r->res_ls, "receive_convert_reply %x error %d",
- lkb->lkb_id, ms->m_result);
+ log_error(r->res_ls, "receive_convert_reply %x remote %d %x %d",
+ lkb->lkb_id, ms->m_header.h_nodeid, ms->m_lkid,
+ ms->m_result);
+ dlm_print_rsb(r);
+ dlm_print_lkb(lkb);
}
}
@@ -3661,20 +4660,18 @@ static void _receive_convert_reply(struct dlm_lkb *lkb, struct dlm_message *ms)
put_rsb(r);
}
-static void receive_convert_reply(struct dlm_ls *ls, struct dlm_message *ms)
+static int receive_convert_reply(struct dlm_ls *ls, struct dlm_message *ms)
{
struct dlm_lkb *lkb;
int error;
error = find_lkb(ls, ms->m_remid, &lkb);
- if (error) {
- log_debug(ls, "receive_convert_reply from %d no lkb %x",
- ms->m_header.h_nodeid, ms->m_remid);
- return;
- }
+ if (error)
+ return error;
_receive_convert_reply(lkb, ms);
dlm_put_lkb(lkb);
+ return 0;
}
static void _receive_unlock_reply(struct dlm_lkb *lkb, struct dlm_message *ms)
@@ -3713,20 +4710,18 @@ static void _receive_unlock_reply(struct dlm_lkb *lkb, struct dlm_message *ms)
put_rsb(r);
}
-static void receive_unlock_reply(struct dlm_ls *ls, struct dlm_message *ms)
+static int receive_unlock_reply(struct dlm_ls *ls, struct dlm_message *ms)
{
struct dlm_lkb *lkb;
int error;
error = find_lkb(ls, ms->m_remid, &lkb);
- if (error) {
- log_debug(ls, "receive_unlock_reply from %d no lkb %x",
- ms->m_header.h_nodeid, ms->m_remid);
- return;
- }
+ if (error)
+ return error;
_receive_unlock_reply(lkb, ms);
dlm_put_lkb(lkb);
+ return 0;
}
static void _receive_cancel_reply(struct dlm_lkb *lkb, struct dlm_message *ms)
@@ -3765,20 +4760,18 @@ static void _receive_cancel_reply(struct dlm_lkb *lkb, struct dlm_message *ms)
put_rsb(r);
}
-static void receive_cancel_reply(struct dlm_ls *ls, struct dlm_message *ms)
+static int receive_cancel_reply(struct dlm_ls *ls, struct dlm_message *ms)
{
struct dlm_lkb *lkb;
int error;
error = find_lkb(ls, ms->m_remid, &lkb);
- if (error) {
- log_debug(ls, "receive_cancel_reply from %d no lkb %x",
- ms->m_header.h_nodeid, ms->m_remid);
- return;
- }
+ if (error)
+ return error;
_receive_cancel_reply(lkb, ms);
dlm_put_lkb(lkb);
+ return 0;
}
static void receive_lookup_reply(struct dlm_ls *ls, struct dlm_message *ms)
@@ -3786,14 +4779,15 @@ static void receive_lookup_reply(struct dlm_ls *ls, struct dlm_message *ms)
struct dlm_lkb *lkb;
struct dlm_rsb *r;
int error, ret_nodeid;
+ int do_lookup_list = 0;
error = find_lkb(ls, ms->m_lkid, &lkb);
if (error) {
- log_error(ls, "receive_lookup_reply no lkb");
+ log_error(ls, "receive_lookup_reply no lkid %x", ms->m_lkid);
return;
}
- /* ms->m_result is the value returned by dlm_dir_lookup on dir node
+ /* ms->m_result is the value returned by dlm_master_lookup on dir node
FIXME: will a non-zero error ever be returned? */
r = lkb->lkb_resource;
@@ -3805,12 +4799,37 @@ static void receive_lookup_reply(struct dlm_ls *ls, struct dlm_message *ms)
goto out;
ret_nodeid = ms->m_nodeid;
+
+ /* We sometimes receive a request from the dir node for this
+ rsb before we've received the dir node's loookup_reply for it.
+ The request from the dir node implies we're the master, so we set
+ ourself as master in receive_request_reply, and verify here that
+ we are indeed the master. */
+
+ if (r->res_master_nodeid && (r->res_master_nodeid != ret_nodeid)) {
+ /* This should never happen */
+ log_error(ls, "receive_lookup_reply %x from %d ret %d "
+ "master %d dir %d our %d first %x %s",
+ lkb->lkb_id, ms->m_header.h_nodeid, ret_nodeid,
+ r->res_master_nodeid, r->res_dir_nodeid,
+ dlm_our_nodeid(), r->res_first_lkid, r->res_name);
+ }
+
if (ret_nodeid == dlm_our_nodeid()) {
+ r->res_master_nodeid = ret_nodeid;
r->res_nodeid = 0;
- ret_nodeid = 0;
+ do_lookup_list = 1;
r->res_first_lkid = 0;
+ } else if (ret_nodeid == -1) {
+ /* the remote node doesn't believe it's the dir node */
+ log_error(ls, "receive_lookup_reply %x from %d bad ret_nodeid",
+ lkb->lkb_id, ms->m_header.h_nodeid);
+ r->res_master_nodeid = 0;
+ r->res_nodeid = -1;
+ lkb->lkb_nodeid = -1;
} else {
- /* set_master() will copy res_nodeid to lkb_nodeid */
+ /* set_master() will set lkb_nodeid from r */
+ r->res_master_nodeid = ret_nodeid;
r->res_nodeid = ret_nodeid;
}
@@ -3825,7 +4844,7 @@ static void receive_lookup_reply(struct dlm_ls *ls, struct dlm_message *ms)
_request_lock(r, lkb);
out_list:
- if (!ret_nodeid)
+ if (do_lookup_list)
process_lookup_list(r);
out:
unlock_rsb(r);
@@ -3833,10 +4852,13 @@ static void receive_lookup_reply(struct dlm_ls *ls, struct dlm_message *ms)
dlm_put_lkb(lkb);
}
-static void _receive_message(struct dlm_ls *ls, struct dlm_message *ms)
+static void _receive_message(struct dlm_ls *ls, struct dlm_message *ms,
+ uint32_t saved_seq)
{
+ int error = 0, noent = 0;
+
if (!dlm_is_member(ls, ms->m_header.h_nodeid)) {
- log_debug(ls, "ignore non-member message %d from %d %x %x %d",
+ log_limit(ls, "receive %d from non-member %d %x %x %d",
ms->m_type, ms->m_header.h_nodeid, ms->m_lkid,
ms->m_remid, ms->m_result);
return;
@@ -3847,47 +4869,50 @@ static void _receive_message(struct dlm_ls *ls, struct dlm_message *ms)
/* messages sent to a master node */
case DLM_MSG_REQUEST:
- receive_request(ls, ms);
+ error = receive_request(ls, ms);
break;
case DLM_MSG_CONVERT:
- receive_convert(ls, ms);
+ error = receive_convert(ls, ms);
break;
case DLM_MSG_UNLOCK:
- receive_unlock(ls, ms);
+ error = receive_unlock(ls, ms);
break;
case DLM_MSG_CANCEL:
- receive_cancel(ls, ms);
+ noent = 1;
+ error = receive_cancel(ls, ms);
break;
/* messages sent from a master node (replies to above) */
case DLM_MSG_REQUEST_REPLY:
- receive_request_reply(ls, ms);
+ error = receive_request_reply(ls, ms);
break;
case DLM_MSG_CONVERT_REPLY:
- receive_convert_reply(ls, ms);
+ error = receive_convert_reply(ls, ms);
break;
case DLM_MSG_UNLOCK_REPLY:
- receive_unlock_reply(ls, ms);
+ error = receive_unlock_reply(ls, ms);
break;
case DLM_MSG_CANCEL_REPLY:
- receive_cancel_reply(ls, ms);
+ error = receive_cancel_reply(ls, ms);
break;
/* messages sent from a master node (only two types of async msg) */
case DLM_MSG_GRANT:
- receive_grant(ls, ms);
+ noent = 1;
+ error = receive_grant(ls, ms);
break;
case DLM_MSG_BAST:
- receive_bast(ls, ms);
+ noent = 1;
+ error = receive_bast(ls, ms);
break;
/* messages sent to a dir node */
@@ -3916,7 +4941,36 @@ static void _receive_message(struct dlm_ls *ls, struct dlm_message *ms)
log_error(ls, "unknown message type %d", ms->m_type);
}
- dlm_astd_wake();
+ /*
+ * When checking for ENOENT, we're checking the result of
+ * find_lkb(m_remid):
+ *
+ * The lock id referenced in the message wasn't found. This may
+ * happen in normal usage for the async messages and cancel, so
+ * only use log_debug for them.
+ *
+ * Some errors are expected and normal.
+ */
+
+ if (error == -ENOENT && noent) {
+ log_debug(ls, "receive %d no %x remote %d %x saved_seq %u",
+ ms->m_type, ms->m_remid, ms->m_header.h_nodeid,
+ ms->m_lkid, saved_seq);
+ } else if (error == -ENOENT) {
+ log_error(ls, "receive %d no %x remote %d %x saved_seq %u",
+ ms->m_type, ms->m_remid, ms->m_header.h_nodeid,
+ ms->m_lkid, saved_seq);
+
+ if (ms->m_type == DLM_MSG_CONVERT)
+ dlm_dump_rsb_hash(ls, ms->m_hash);
+ }
+
+ if (error == -EINVAL) {
+ log_error(ls, "receive %d inval from %d lkid %x remid %x "
+ "saved_seq %u",
+ ms->m_type, ms->m_header.h_nodeid,
+ ms->m_lkid, ms->m_remid, saved_seq);
+ }
}
/* If the lockspace is in recovery mode (locking stopped), then normal
@@ -3931,19 +4985,29 @@ static void dlm_receive_message(struct dlm_ls *ls, struct dlm_message *ms,
int nodeid)
{
if (dlm_locking_stopped(ls)) {
+ /* If we were a member of this lockspace, left, and rejoined,
+ other nodes may still be sending us messages from the
+ lockspace generation before we left. */
+ if (!ls->ls_generation) {
+ log_limit(ls, "receive %d from %d ignore old gen",
+ ms->m_type, nodeid);
+ return;
+ }
+
dlm_add_requestqueue(ls, nodeid, ms);
} else {
dlm_wait_requestqueue(ls);
- _receive_message(ls, ms);
+ _receive_message(ls, ms, 0);
}
}
/* This is called by dlm_recoverd to process messages that were saved on
the requestqueue. */
-void dlm_receive_message_saved(struct dlm_ls *ls, struct dlm_message *ms)
+void dlm_receive_message_saved(struct dlm_ls *ls, struct dlm_message *ms,
+ uint32_t saved_seq)
{
- _receive_message(ls, ms);
+ _receive_message(ls, ms, saved_seq);
}
/* This is called by the midcomms layer when something is received for
@@ -3979,9 +5043,11 @@ void dlm_receive_buffer(union dlm_packet *p, int nodeid)
ls = dlm_find_lockspace_global(hd->h_lockspace);
if (!ls) {
- if (dlm_config.ci_log_debug)
- log_print("invalid lockspace %x from %d cmd %d type %d",
- hd->h_lockspace, nodeid, hd->h_cmd, type);
+ if (dlm_config.ci_log_debug) {
+ printk_ratelimited(KERN_DEBUG "dlm: invalid lockspace "
+ "%u from %d cmd %d type %d\n",
+ hd->h_lockspace, nodeid, hd->h_cmd, type);
+ }
if (hd->h_cmd == DLM_RCOM && type == DLM_RCOM_STATUS)
dlm_send_ls_not_ready(nodeid, &p->rcom);
@@ -4001,15 +5067,17 @@ void dlm_receive_buffer(union dlm_packet *p, int nodeid)
dlm_put_lockspace(ls);
}
-static void recover_convert_waiter(struct dlm_ls *ls, struct dlm_lkb *lkb)
+static void recover_convert_waiter(struct dlm_ls *ls, struct dlm_lkb *lkb,
+ struct dlm_message *ms_stub)
{
if (middle_conversion(lkb)) {
hold_lkb(lkb);
- ls->ls_stub_ms.m_type = DLM_MSG_CONVERT_REPLY;
- ls->ls_stub_ms.m_result = -EINPROGRESS;
- ls->ls_stub_ms.m_flags = lkb->lkb_flags;
- ls->ls_stub_ms.m_header.h_nodeid = lkb->lkb_nodeid;
- _receive_convert_reply(lkb, &ls->ls_stub_ms);
+ memset(ms_stub, 0, sizeof(struct dlm_message));
+ ms_stub->m_flags = DLM_IFL_STUB_MS;
+ ms_stub->m_type = DLM_MSG_CONVERT_REPLY;
+ ms_stub->m_result = -EINPROGRESS;
+ ms_stub->m_header.h_nodeid = lkb->lkb_nodeid;
+ _receive_convert_reply(lkb, ms_stub);
/* Same special case as in receive_rcom_lock_args() */
lkb->lkb_grmode = DLM_LOCK_IV;
@@ -4027,15 +5095,13 @@ static void recover_convert_waiter(struct dlm_ls *ls, struct dlm_lkb *lkb)
/* A waiting lkb needs recovery if the master node has failed, or
the master node is changing (only when no directory is used) */
-static int waiter_needs_recovery(struct dlm_ls *ls, struct dlm_lkb *lkb)
+static int waiter_needs_recovery(struct dlm_ls *ls, struct dlm_lkb *lkb,
+ int dir_nodeid)
{
- if (dlm_is_removed(ls, lkb->lkb_nodeid))
+ if (dlm_no_directory(ls))
return 1;
- if (!dlm_no_directory(ls))
- return 0;
-
- if (dlm_dir_nodeid(lkb->lkb_resource) != lkb->lkb_nodeid)
+ if (dlm_is_removed(ls, lkb->lkb_wait_nodeid))
return 1;
return 0;
@@ -4050,13 +5116,36 @@ static int waiter_needs_recovery(struct dlm_ls *ls, struct dlm_lkb *lkb)
void dlm_recover_waiters_pre(struct dlm_ls *ls)
{
struct dlm_lkb *lkb, *safe;
+ struct dlm_message *ms_stub;
int wait_type, stub_unlock_result, stub_cancel_result;
+ int dir_nodeid;
+
+ ms_stub = kmalloc(sizeof(struct dlm_message), GFP_KERNEL);
+ if (!ms_stub) {
+ log_error(ls, "dlm_recover_waiters_pre no mem");
+ return;
+ }
mutex_lock(&ls->ls_waiters_mutex);
list_for_each_entry_safe(lkb, safe, &ls->ls_waiters, lkb_wait_reply) {
- log_debug(ls, "pre recover waiter lkid %x type %d flags %x",
- lkb->lkb_id, lkb->lkb_wait_type, lkb->lkb_flags);
+
+ dir_nodeid = dlm_dir_nodeid(lkb->lkb_resource);
+
+ /* exclude debug messages about unlocks because there can be so
+ many and they aren't very interesting */
+
+ if (lkb->lkb_wait_type != DLM_MSG_UNLOCK) {
+ log_debug(ls, "waiter %x remote %x msg %d r_nodeid %d "
+ "lkb_nodeid %d wait_nodeid %d dir_nodeid %d",
+ lkb->lkb_id,
+ lkb->lkb_remid,
+ lkb->lkb_wait_type,
+ lkb->lkb_resource->res_nodeid,
+ lkb->lkb_nodeid,
+ lkb->lkb_wait_nodeid,
+ dir_nodeid);
+ }
/* all outstanding lookups, regardless of destination will be
resent after recovery is done */
@@ -4066,7 +5155,7 @@ void dlm_recover_waiters_pre(struct dlm_ls *ls)
continue;
}
- if (!waiter_needs_recovery(ls, lkb))
+ if (!waiter_needs_recovery(ls, lkb, dir_nodeid))
continue;
wait_type = lkb->lkb_wait_type;
@@ -4102,26 +5191,28 @@ void dlm_recover_waiters_pre(struct dlm_ls *ls)
break;
case DLM_MSG_CONVERT:
- recover_convert_waiter(ls, lkb);
+ recover_convert_waiter(ls, lkb, ms_stub);
break;
case DLM_MSG_UNLOCK:
hold_lkb(lkb);
- ls->ls_stub_ms.m_type = DLM_MSG_UNLOCK_REPLY;
- ls->ls_stub_ms.m_result = stub_unlock_result;
- ls->ls_stub_ms.m_flags = lkb->lkb_flags;
- ls->ls_stub_ms.m_header.h_nodeid = lkb->lkb_nodeid;
- _receive_unlock_reply(lkb, &ls->ls_stub_ms);
+ memset(ms_stub, 0, sizeof(struct dlm_message));
+ ms_stub->m_flags = DLM_IFL_STUB_MS;
+ ms_stub->m_type = DLM_MSG_UNLOCK_REPLY;
+ ms_stub->m_result = stub_unlock_result;
+ ms_stub->m_header.h_nodeid = lkb->lkb_nodeid;
+ _receive_unlock_reply(lkb, ms_stub);
dlm_put_lkb(lkb);
break;
case DLM_MSG_CANCEL:
hold_lkb(lkb);
- ls->ls_stub_ms.m_type = DLM_MSG_CANCEL_REPLY;
- ls->ls_stub_ms.m_result = stub_cancel_result;
- ls->ls_stub_ms.m_flags = lkb->lkb_flags;
- ls->ls_stub_ms.m_header.h_nodeid = lkb->lkb_nodeid;
- _receive_cancel_reply(lkb, &ls->ls_stub_ms);
+ memset(ms_stub, 0, sizeof(struct dlm_message));
+ ms_stub->m_flags = DLM_IFL_STUB_MS;
+ ms_stub->m_type = DLM_MSG_CANCEL_REPLY;
+ ms_stub->m_result = stub_cancel_result;
+ ms_stub->m_header.h_nodeid = lkb->lkb_nodeid;
+ _receive_cancel_reply(lkb, ms_stub);
dlm_put_lkb(lkb);
break;
@@ -4132,6 +5223,7 @@ void dlm_recover_waiters_pre(struct dlm_ls *ls)
schedule();
}
mutex_unlock(&ls->ls_waiters_mutex);
+ kfree(ms_stub);
}
static struct dlm_lkb *find_resend_waiter(struct dlm_ls *ls)
@@ -4196,8 +5288,11 @@ int dlm_recover_waiters_post(struct dlm_ls *ls)
ou = is_overlap_unlock(lkb);
err = 0;
- log_debug(ls, "recover_waiters_post %x type %d flags %x %s",
- lkb->lkb_id, mstype, lkb->lkb_flags, r->res_name);
+ log_debug(ls, "waiter %x remote %x msg %d r_nodeid %d "
+ "lkb_nodeid %d wait_nodeid %d dir_nodeid %d "
+ "overlap %d %d", lkb->lkb_id, lkb->lkb_remid, mstype,
+ r->res_nodeid, lkb->lkb_nodeid, lkb->lkb_wait_nodeid,
+ dlm_dir_nodeid(r), oc, ou);
/* At this point we assume that we won't get a reply to any
previous op or overlap op on this lock. First, do a big
@@ -4249,9 +5344,12 @@ int dlm_recover_waiters_post(struct dlm_ls *ls)
}
}
- if (err)
- log_error(ls, "recover_waiters_post %x %d %x %d %d",
- lkb->lkb_id, mstype, lkb->lkb_flags, oc, ou);
+ if (err) {
+ log_error(ls, "waiter %x msg %d r_nodeid %d "
+ "dir_nodeid %d overlap %d %d",
+ lkb->lkb_id, mstype, r->res_nodeid,
+ dlm_dir_nodeid(r), oc, ou);
+ }
unlock_rsb(r);
put_rsb(r);
dlm_put_lkb(lkb);
@@ -4260,110 +5358,187 @@ int dlm_recover_waiters_post(struct dlm_ls *ls)
return error;
}
-static void purge_queue(struct dlm_rsb *r, struct list_head *queue,
- int (*test)(struct dlm_ls *ls, struct dlm_lkb *lkb))
+static void purge_mstcpy_list(struct dlm_ls *ls, struct dlm_rsb *r,
+ struct list_head *list)
{
- struct dlm_ls *ls = r->res_ls;
struct dlm_lkb *lkb, *safe;
- list_for_each_entry_safe(lkb, safe, queue, lkb_statequeue) {
- if (test(ls, lkb)) {
- rsb_set_flag(r, RSB_LOCKS_PURGED);
- del_lkb(r, lkb);
- /* this put should free the lkb */
- if (!dlm_put_lkb(lkb))
- log_error(ls, "purged lkb not released");
- }
+ list_for_each_entry_safe(lkb, safe, list, lkb_statequeue) {
+ if (!is_master_copy(lkb))
+ continue;
+
+ /* don't purge lkbs we've added in recover_master_copy for
+ the current recovery seq */
+
+ if (lkb->lkb_recover_seq == ls->ls_recover_seq)
+ continue;
+
+ del_lkb(r, lkb);
+
+ /* this put should free the lkb */
+ if (!dlm_put_lkb(lkb))
+ log_error(ls, "purged mstcpy lkb not released");
}
}
-static int purge_dead_test(struct dlm_ls *ls, struct dlm_lkb *lkb)
+void dlm_purge_mstcpy_locks(struct dlm_rsb *r)
{
- return (is_master_copy(lkb) && dlm_is_removed(ls, lkb->lkb_nodeid));
-}
+ struct dlm_ls *ls = r->res_ls;
-static int purge_mstcpy_test(struct dlm_ls *ls, struct dlm_lkb *lkb)
-{
- return is_master_copy(lkb);
+ purge_mstcpy_list(ls, r, &r->res_grantqueue);
+ purge_mstcpy_list(ls, r, &r->res_convertqueue);
+ purge_mstcpy_list(ls, r, &r->res_waitqueue);
}
-static void purge_dead_locks(struct dlm_rsb *r)
+static void purge_dead_list(struct dlm_ls *ls, struct dlm_rsb *r,
+ struct list_head *list,
+ int nodeid_gone, unsigned int *count)
{
- purge_queue(r, &r->res_grantqueue, &purge_dead_test);
- purge_queue(r, &r->res_convertqueue, &purge_dead_test);
- purge_queue(r, &r->res_waitqueue, &purge_dead_test);
-}
+ struct dlm_lkb *lkb, *safe;
-void dlm_purge_mstcpy_locks(struct dlm_rsb *r)
-{
- purge_queue(r, &r->res_grantqueue, &purge_mstcpy_test);
- purge_queue(r, &r->res_convertqueue, &purge_mstcpy_test);
- purge_queue(r, &r->res_waitqueue, &purge_mstcpy_test);
+ list_for_each_entry_safe(lkb, safe, list, lkb_statequeue) {
+ if (!is_master_copy(lkb))
+ continue;
+
+ if ((lkb->lkb_nodeid == nodeid_gone) ||
+ dlm_is_removed(ls, lkb->lkb_nodeid)) {
+
+ /* tell recover_lvb to invalidate the lvb
+ because a node holding EX/PW failed */
+ if ((lkb->lkb_exflags & DLM_LKF_VALBLK) &&
+ (lkb->lkb_grmode >= DLM_LOCK_PW)) {
+ rsb_set_flag(r, RSB_RECOVER_LVB_INVAL);
+ }
+
+ del_lkb(r, lkb);
+
+ /* this put should free the lkb */
+ if (!dlm_put_lkb(lkb))
+ log_error(ls, "purged dead lkb not released");
+
+ rsb_set_flag(r, RSB_RECOVER_GRANT);
+
+ (*count)++;
+ }
+ }
}
/* Get rid of locks held by nodes that are gone. */
-int dlm_purge_locks(struct dlm_ls *ls)
+void dlm_recover_purge(struct dlm_ls *ls)
{
struct dlm_rsb *r;
+ struct dlm_member *memb;
+ int nodes_count = 0;
+ int nodeid_gone = 0;
+ unsigned int lkb_count = 0;
+
+ /* cache one removed nodeid to optimize the common
+ case of a single node removed */
- log_debug(ls, "dlm_purge_locks");
+ list_for_each_entry(memb, &ls->ls_nodes_gone, list) {
+ nodes_count++;
+ nodeid_gone = memb->nodeid;
+ }
+
+ if (!nodes_count)
+ return;
down_write(&ls->ls_root_sem);
list_for_each_entry(r, &ls->ls_root_list, res_root_list) {
hold_rsb(r);
lock_rsb(r);
- if (is_master(r))
- purge_dead_locks(r);
+ if (is_master(r)) {
+ purge_dead_list(ls, r, &r->res_grantqueue,
+ nodeid_gone, &lkb_count);
+ purge_dead_list(ls, r, &r->res_convertqueue,
+ nodeid_gone, &lkb_count);
+ purge_dead_list(ls, r, &r->res_waitqueue,
+ nodeid_gone, &lkb_count);
+ }
unlock_rsb(r);
unhold_rsb(r);
-
- schedule();
+ cond_resched();
}
up_write(&ls->ls_root_sem);
- return 0;
+ if (lkb_count)
+ log_rinfo(ls, "dlm_recover_purge %u locks for %u nodes",
+ lkb_count, nodes_count);
}
-static struct dlm_rsb *find_purged_rsb(struct dlm_ls *ls, int bucket)
+static struct dlm_rsb *find_grant_rsb(struct dlm_ls *ls, int bucket)
{
- struct dlm_rsb *r, *r_ret = NULL;
+ struct rb_node *n;
+ struct dlm_rsb *r;
spin_lock(&ls->ls_rsbtbl[bucket].lock);
- list_for_each_entry(r, &ls->ls_rsbtbl[bucket].list, res_hashchain) {
- if (!rsb_flag(r, RSB_LOCKS_PURGED))
+ for (n = rb_first(&ls->ls_rsbtbl[bucket].keep); n; n = rb_next(n)) {
+ r = rb_entry(n, struct dlm_rsb, res_hashnode);
+
+ if (!rsb_flag(r, RSB_RECOVER_GRANT))
+ continue;
+ if (!is_master(r)) {
+ rsb_clear_flag(r, RSB_RECOVER_GRANT);
continue;
+ }
hold_rsb(r);
- rsb_clear_flag(r, RSB_LOCKS_PURGED);
- r_ret = r;
- break;
+ spin_unlock(&ls->ls_rsbtbl[bucket].lock);
+ return r;
}
spin_unlock(&ls->ls_rsbtbl[bucket].lock);
- return r_ret;
+ return NULL;
}
-void dlm_grant_after_purge(struct dlm_ls *ls)
+/*
+ * Attempt to grant locks on resources that we are the master of.
+ * Locks may have become grantable during recovery because locks
+ * from departed nodes have been purged (or not rebuilt), allowing
+ * previously blocked locks to now be granted. The subset of rsb's
+ * we are interested in are those with lkb's on either the convert or
+ * waiting queues.
+ *
+ * Simplest would be to go through each master rsb and check for non-empty
+ * convert or waiting queues, and attempt to grant on those rsbs.
+ * Checking the queues requires lock_rsb, though, for which we'd need
+ * to release the rsbtbl lock. This would make iterating through all
+ * rsb's very inefficient. So, we rely on earlier recovery routines
+ * to set RECOVER_GRANT on any rsb's that we should attempt to grant
+ * locks for.
+ */
+
+void dlm_recover_grant(struct dlm_ls *ls)
{
struct dlm_rsb *r;
int bucket = 0;
+ unsigned int count = 0;
+ unsigned int rsb_count = 0;
+ unsigned int lkb_count = 0;
while (1) {
- r = find_purged_rsb(ls, bucket);
+ r = find_grant_rsb(ls, bucket);
if (!r) {
if (bucket == ls->ls_rsbtbl_size - 1)
break;
bucket++;
continue;
}
+ rsb_count++;
+ count = 0;
lock_rsb(r);
- if (is_master(r)) {
- grant_pending_locks(r);
- confirm_master(r, 0);
- }
+ /* the RECOVER_GRANT flag is checked in the grant path */
+ grant_pending_locks(r, &count);
+ rsb_clear_flag(r, RSB_RECOVER_GRANT);
+ lkb_count += count;
+ confirm_master(r, 0);
unlock_rsb(r);
put_rsb(r);
- schedule();
+ cond_resched();
}
+
+ if (lkb_count)
+ log_rinfo(ls, "dlm_recover_grant %u locks on %u resources",
+ lkb_count, rsb_count);
}
static struct dlm_lkb *search_remid_list(struct list_head *head, int nodeid,
@@ -4412,8 +5587,8 @@ static int receive_rcom_lock_args(struct dlm_ls *ls, struct dlm_lkb *lkb,
lkb->lkb_grmode = rl->rl_grmode;
/* don't set lkb_status because add_lkb wants to itself */
- lkb->lkb_bastfn = (rl->rl_asts & AST_BAST) ? &fake_bastfn : NULL;
- lkb->lkb_astfn = (rl->rl_asts & AST_COMP) ? &fake_astfn : NULL;
+ lkb->lkb_bastfn = (rl->rl_asts & DLM_CB_BAST) ? &fake_bastfn : NULL;
+ lkb->lkb_astfn = (rl->rl_asts & DLM_CB_CAST) ? &fake_astfn : NULL;
if (lkb->lkb_exflags & DLM_LKF_VALBLK) {
int lvblen = rc->rc_header.h_length - sizeof(struct dlm_rcom) -
@@ -4452,6 +5627,8 @@ int dlm_recover_master_copy(struct dlm_ls *ls, struct dlm_rcom *rc)
struct rcom_lock *rl = (struct rcom_lock *) rc->rc_buf;
struct dlm_rsb *r;
struct dlm_lkb *lkb;
+ uint32_t remid = 0;
+ int from_nodeid = rc->rc_header.h_nodeid;
int error;
if (rl->rl_parent_lkid) {
@@ -4459,14 +5636,31 @@ int dlm_recover_master_copy(struct dlm_ls *ls, struct dlm_rcom *rc)
goto out;
}
+ remid = le32_to_cpu(rl->rl_lkid);
+
+ /* In general we expect the rsb returned to be R_MASTER, but we don't
+ have to require it. Recovery of masters on one node can overlap
+ recovery of locks on another node, so one node can send us MSTCPY
+ locks before we've made ourselves master of this rsb. We can still
+ add new MSTCPY locks that we receive here without any harm; when
+ we make ourselves master, dlm_recover_masters() won't touch the
+ MSTCPY locks we've received early. */
+
error = find_rsb(ls, rl->rl_name, le16_to_cpu(rl->rl_namelen),
- R_MASTER, &r);
+ from_nodeid, R_RECEIVE_RECOVER, &r);
if (error)
goto out;
lock_rsb(r);
- lkb = search_remid(r, rc->rc_header.h_nodeid, le32_to_cpu(rl->rl_lkid));
+ if (dlm_no_directory(ls) && (dlm_dir_nodeid(r) != dlm_our_nodeid())) {
+ log_error(ls, "dlm_recover_master_copy remote %d %x not dir",
+ from_nodeid, remid);
+ error = -EBADR;
+ goto out_unlock;
+ }
+
+ lkb = search_remid(r, from_nodeid, remid);
if (lkb) {
error = -EEXIST;
goto out_remid;
@@ -4485,19 +5679,25 @@ int dlm_recover_master_copy(struct dlm_ls *ls, struct dlm_rcom *rc)
attach_lkb(r, lkb);
add_lkb(r, lkb, rl->rl_status);
error = 0;
+ ls->ls_recover_locks_in++;
+
+ if (!list_empty(&r->res_waitqueue) || !list_empty(&r->res_convertqueue))
+ rsb_set_flag(r, RSB_RECOVER_GRANT);
out_remid:
/* this is the new value returned to the lock holder for
saving in its process-copy lkb */
rl->rl_remid = cpu_to_le32(lkb->lkb_id);
+ lkb->lkb_recover_seq = ls->ls_recover_seq;
+
out_unlock:
unlock_rsb(r);
put_rsb(r);
out:
- if (error)
- log_debug(ls, "recover_master_copy %d %x", error,
- le32_to_cpu(rl->rl_lkid));
+ if (error && error != -EEXIST)
+ log_rinfo(ls, "dlm_recover_master_copy remote %d %x error %d",
+ from_nodeid, remid, error);
rl->rl_result = cpu_to_le32(error);
return error;
}
@@ -4508,41 +5708,52 @@ int dlm_recover_process_copy(struct dlm_ls *ls, struct dlm_rcom *rc)
struct rcom_lock *rl = (struct rcom_lock *) rc->rc_buf;
struct dlm_rsb *r;
struct dlm_lkb *lkb;
- int error;
+ uint32_t lkid, remid;
+ int error, result;
+
+ lkid = le32_to_cpu(rl->rl_lkid);
+ remid = le32_to_cpu(rl->rl_remid);
+ result = le32_to_cpu(rl->rl_result);
- error = find_lkb(ls, le32_to_cpu(rl->rl_lkid), &lkb);
+ error = find_lkb(ls, lkid, &lkb);
if (error) {
- log_error(ls, "recover_process_copy no lkid %x",
- le32_to_cpu(rl->rl_lkid));
+ log_error(ls, "dlm_recover_process_copy no %x remote %d %x %d",
+ lkid, rc->rc_header.h_nodeid, remid, result);
return error;
}
- DLM_ASSERT(is_process_copy(lkb), dlm_print_lkb(lkb););
-
- error = le32_to_cpu(rl->rl_result);
-
r = lkb->lkb_resource;
hold_rsb(r);
lock_rsb(r);
- switch (error) {
+ if (!is_process_copy(lkb)) {
+ log_error(ls, "dlm_recover_process_copy bad %x remote %d %x %d",
+ lkid, rc->rc_header.h_nodeid, remid, result);
+ dlm_dump_rsb(r);
+ unlock_rsb(r);
+ put_rsb(r);
+ dlm_put_lkb(lkb);
+ return -EINVAL;
+ }
+
+ switch (result) {
case -EBADR:
/* There's a chance the new master received our lock before
dlm_recover_master_reply(), this wouldn't happen if we did
a barrier between recover_masters and recover_locks. */
- log_debug(ls, "master copy not ready %x r %lx %s", lkb->lkb_id,
- (unsigned long)r, r->res_name);
+
+ log_debug(ls, "dlm_recover_process_copy %x remote %d %x %d",
+ lkid, rc->rc_header.h_nodeid, remid, result);
+
dlm_send_rcom_lock(r, lkb);
goto out;
case -EEXIST:
- log_debug(ls, "master copy exists %x", lkb->lkb_id);
- /* fall through */
case 0:
- lkb->lkb_remid = le32_to_cpu(rl->rl_remid);
+ lkb->lkb_remid = remid;
break;
default:
- log_error(ls, "dlm_recover_process_copy unknown error %d %x",
- error, lkb->lkb_id);
+ log_error(ls, "dlm_recover_process_copy %x remote %d %x %d unk",
+ lkid, rc->rc_header.h_nodeid, remid, result);
}
/* an ack for dlm_recover_locks() which waits for replies from
@@ -4589,7 +5800,6 @@ int dlm_user_request(struct dlm_ls *ls, struct dlm_user_args *ua,
error = set_lock_args(mode, &ua->lksb, flags, namelen, timeout_cs,
fake_astfn, ua, fake_bastfn, &args);
lkb->lkb_flags |= DLM_IFL_USER;
- ua->old_mode = DLM_LOCK_IV;
if (error) {
__put_lkb(ls, lkb);
@@ -4658,7 +5868,6 @@ int dlm_user_convert(struct dlm_ls *ls, struct dlm_user_args *ua_tmp,
ua->bastparam = ua_tmp->bastparam;
ua->bastaddr = ua_tmp->bastaddr;
ua->user_lksb = ua_tmp->user_lksb;
- ua->old_mode = lkb->lkb_grmode;
error = set_lock_args(mode, &ua->lksb, flags, 0, timeout_cs,
fake_astfn, ua, fake_bastfn, &args);
@@ -4714,7 +5923,7 @@ int dlm_user_unlock(struct dlm_ls *ls, struct dlm_user_args *ua_tmp,
goto out_put;
spin_lock(&ua->proc->locks_spin);
- /* dlm_user_add_ast() may have already taken lkb off the proc list */
+ /* dlm_user_add_cb() may have already taken lkb off the proc list */
if (!list_empty(&lkb->lkb_ownqueue))
list_move(&lkb->lkb_ownqueue, &ua->proc->unlocking);
spin_unlock(&ua->proc->locks_spin);
@@ -4833,15 +6042,18 @@ static int orphan_proc_lock(struct dlm_ls *ls, struct dlm_lkb *lkb)
return error;
}
-/* The force flag allows the unlock to go ahead even if the lkb isn't granted.
- Regardless of what rsb queue the lock is on, it's removed and freed. */
+/* The FORCEUNLOCK flag allows the unlock to go ahead even if the lkb isn't
+ granted. Regardless of what rsb queue the lock is on, it's removed and
+ freed. The IVVALBLK flag causes the lvb on the resource to be invalidated
+ if our lock is PW/EX (it's ignored if our granted mode is smaller.) */
static int unlock_proc_lock(struct dlm_ls *ls, struct dlm_lkb *lkb)
{
struct dlm_args args;
int error;
- set_unlock_args(DLM_LKF_FORCEUNLOCK, lkb->lkb_ua, &args);
+ set_unlock_args(DLM_LKF_FORCEUNLOCK | DLM_LKF_IVVALBLK,
+ lkb->lkb_ua, &args);
error = unlock_lock(ls, lkb, &args);
if (error == -DLM_EUNLOCK)
@@ -4851,7 +6063,7 @@ static int unlock_proc_lock(struct dlm_ls *ls, struct dlm_lkb *lkb)
/* We have to release clear_proc_locks mutex before calling unlock_proc_lock()
(which does lock_rsb) due to deadlock with receiving a message that does
- lock_rsb followed by dlm_user_add_ast() */
+ lock_rsb followed by dlm_user_add_cb() */
static struct dlm_lkb *del_proc_lock(struct dlm_ls *ls,
struct dlm_user_proc *proc)
@@ -4874,7 +6086,7 @@ static struct dlm_lkb *del_proc_lock(struct dlm_ls *ls,
return lkb;
}
-/* The ls_clear_proc_locks mutex protects against dlm_user_add_asts() which
+/* The ls_clear_proc_locks mutex protects against dlm_user_add_cb() which
1) references lkb->ua which we free here and 2) adds lkbs to proc->asts,
which we clear here. */
@@ -4916,9 +6128,10 @@ void dlm_clear_proc_locks(struct dlm_ls *ls, struct dlm_user_proc *proc)
dlm_put_lkb(lkb);
}
- list_for_each_entry_safe(lkb, safe, &proc->asts, lkb_astqueue) {
- lkb->lkb_ast_type = 0;
- list_del(&lkb->lkb_astqueue);
+ list_for_each_entry_safe(lkb, safe, &proc->asts, lkb_cb_list) {
+ memset(&lkb->lkb_callbacks, 0,
+ sizeof(struct dlm_callback) * DLM_CALLBACKS_SIZE);
+ list_del_init(&lkb->lkb_cb_list);
dlm_put_lkb(lkb);
}
@@ -4957,8 +6170,10 @@ static void purge_proc_locks(struct dlm_ls *ls, struct dlm_user_proc *proc)
spin_unlock(&proc->locks_spin);
spin_lock(&proc->asts_spin);
- list_for_each_entry_safe(lkb, safe, &proc->asts, lkb_astqueue) {
- list_del(&lkb->lkb_astqueue);
+ list_for_each_entry_safe(lkb, safe, &proc->asts, lkb_cb_list) {
+ memset(&lkb->lkb_callbacks, 0,
+ sizeof(struct dlm_callback) * DLM_CALLBACKS_SIZE);
+ list_del_init(&lkb->lkb_cb_list);
dlm_put_lkb(lkb);
}
spin_unlock(&proc->asts_spin);
diff --git a/fs/dlm/lock.h b/fs/dlm/lock.h
index 88e93c80cc2..5e0c72e36a9 100644
--- a/fs/dlm/lock.h
+++ b/fs/dlm/lock.h
@@ -14,8 +14,10 @@
#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_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, struct dlm_message *ms,
+ uint32_t saved_seq);
void dlm_receive_buffer(union dlm_packet *p, int nodeid);
int dlm_modes_compat(int mode1, int mode2);
void dlm_put_rsb(struct dlm_rsb *r);
@@ -24,12 +26,18 @@ int dlm_put_lkb(struct dlm_lkb *lkb);
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);
+void dlm_scan_waiters(struct dlm_ls *ls);
void dlm_scan_timeout(struct dlm_ls *ls);
void dlm_adjust_timeouts(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_purge_locks(struct dlm_ls *ls);
+int dlm_search_rsb_tree(struct rb_root *tree, char *name, int len,
+ struct dlm_rsb **r_ret);
+
+void dlm_recover_purge(struct dlm_ls *ls);
void dlm_purge_mstcpy_locks(struct dlm_rsb *r);
-void dlm_grant_after_purge(struct dlm_ls *ls);
+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);
diff --git a/fs/dlm/lockspace.c b/fs/dlm/lockspace.c
index f994a7dfda8..f3e72787e7f 100644
--- a/fs/dlm/lockspace.c
+++ b/fs/dlm/lockspace.c
@@ -2,7 +2,7 @@
*******************************************************************************
**
** Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
-** Copyright (C) 2004-2008 Red Hat, Inc. All rights reserved.
+** Copyright (C) 2004-2011 Red Hat, Inc. All rights reserved.
**
** This copyrighted material is made available to anyone wishing to use,
** modify, copy, or redistribute it subject to the terms and conditions
@@ -15,7 +15,6 @@
#include "lockspace.h"
#include "member.h"
#include "recoverd.h"
-#include "ast.h"
#include "dir.h"
#include "lowcomms.h"
#include "config.h"
@@ -24,6 +23,7 @@
#include "recover.h"
#include "requestqueue.h"
#include "user.h"
+#include "ast.h"
static int ls_count;
static struct mutex ls_lock;
@@ -35,8 +35,11 @@ static struct task_struct * scand_task;
static ssize_t dlm_control_store(struct dlm_ls *ls, const char *buf, size_t len)
{
ssize_t ret = len;
- int n = simple_strtol(buf, NULL, 0);
+ int n;
+ int rc = kstrtoint(buf, 0, &n);
+ if (rc)
+ return rc;
ls = dlm_find_lockspace_local(ls->ls_local_handle);
if (!ls)
return -EINVAL;
@@ -57,7 +60,10 @@ static ssize_t dlm_control_store(struct dlm_ls *ls, const char *buf, size_t len)
static ssize_t dlm_event_store(struct dlm_ls *ls, const char *buf, size_t len)
{
- ls->ls_uevent_result = simple_strtol(buf, NULL, 0);
+ int rc = kstrtoint(buf, 0, &ls->ls_uevent_result);
+
+ if (rc)
+ return rc;
set_bit(LSFL_UEVENT_WAIT, &ls->ls_flags);
wake_up(&ls->ls_uevent_wait);
return len;
@@ -70,7 +76,27 @@ static ssize_t dlm_id_show(struct dlm_ls *ls, char *buf)
static ssize_t dlm_id_store(struct dlm_ls *ls, const char *buf, size_t len)
{
- ls->ls_global_id = simple_strtoul(buf, NULL, 0);
+ int rc = kstrtouint(buf, 0, &ls->ls_global_id);
+
+ if (rc)
+ return rc;
+ return len;
+}
+
+static ssize_t dlm_nodir_show(struct dlm_ls *ls, char *buf)
+{
+ return snprintf(buf, PAGE_SIZE, "%u\n", dlm_no_directory(ls));
+}
+
+static ssize_t dlm_nodir_store(struct dlm_ls *ls, const char *buf, size_t len)
+{
+ int val;
+ int rc = kstrtoint(buf, 0, &val);
+
+ if (rc)
+ return rc;
+ if (val == 1)
+ set_bit(LSFL_NODIR, &ls->ls_flags);
return len;
}
@@ -107,6 +133,12 @@ static struct dlm_attr dlm_attr_id = {
.store = dlm_id_store
};
+static struct dlm_attr dlm_attr_nodir = {
+ .attr = {.name = "nodir", .mode = S_IRUGO | S_IWUSR},
+ .show = dlm_nodir_show,
+ .store = dlm_nodir_store
+};
+
static struct dlm_attr dlm_attr_recover_status = {
.attr = {.name = "recover_status", .mode = S_IRUGO},
.show = dlm_recover_status_show
@@ -121,6 +153,7 @@ static struct attribute *dlm_attrs[] = {
&dlm_attr_control.attr,
&dlm_attr_event.attr,
&dlm_attr_id.attr,
+ &dlm_attr_nodir.attr,
&dlm_attr_recover_status.attr,
&dlm_attr_recover_nodeid.attr,
NULL,
@@ -170,7 +203,7 @@ static int do_uevent(struct dlm_ls *ls, int in)
else
kobject_uevent(&ls->ls_kobj, KOBJ_OFFLINE);
- log_debug(ls, "%s the lockspace group...", in ? "joining" : "leaving");
+ log_rinfo(ls, "%s the lockspace group...", in ? "joining" : "leaving");
/* dlm_controld will see the uevent, do the necessary group management
and then write to sysfs to wake us */
@@ -178,7 +211,7 @@ static int do_uevent(struct dlm_ls *ls, int in)
error = wait_event_interruptible(ls->ls_uevent_wait,
test_and_clear_bit(LSFL_UEVENT_WAIT, &ls->ls_flags));
- log_debug(ls, "group event done %d %d", error, ls->ls_uevent_result);
+ log_rinfo(ls, "group event done %d %d", error, ls->ls_uevent_result);
if (error)
goto out;
@@ -243,7 +276,6 @@ static struct dlm_ls *find_ls_to_scan(void)
static int dlm_scand(void *data)
{
struct dlm_ls *ls;
- int timeout_jiffies = dlm_config.ci_scan_secs * HZ;
while (!kthread_should_stop()) {
ls = find_ls_to_scan();
@@ -252,13 +284,14 @@ static int dlm_scand(void *data)
ls->ls_scan_time = jiffies;
dlm_scan_rsbs(ls);
dlm_scan_timeout(ls);
+ dlm_scan_waiters(ls);
dlm_unlock_recovery(ls);
} else {
ls->ls_scan_time += HZ;
}
- } else {
- schedule_timeout_interruptible(timeout_jiffies);
+ continue;
}
+ schedule_timeout_interruptible(dlm_config.ci_scan_secs * HZ);
}
return 0;
}
@@ -359,17 +392,10 @@ static int threads_start(void)
{
int error;
- /* Thread which process lock requests for all lockspace's */
- error = dlm_astd_start();
- if (error) {
- log_print("cannot start dlm_astd thread %d", error);
- goto fail;
- }
-
error = dlm_scand_start();
if (error) {
log_print("cannot start dlm_scand thread %d", error);
- goto astd_fail;
+ goto fail;
}
/* Thread for sending/receiving messages for all lockspace's */
@@ -383,8 +409,6 @@ static int threads_start(void)
scand_fail:
dlm_scand_stop();
- astd_fail:
- dlm_astd_stop();
fail:
return error;
}
@@ -393,15 +417,17 @@ static void threads_stop(void)
{
dlm_scand_stop();
dlm_lowcomms_stop();
- dlm_astd_stop();
}
-static int new_lockspace(const char *name, int namelen, void **lockspace,
- uint32_t flags, int lvblen)
+static int new_lockspace(const char *name, const char *cluster,
+ uint32_t flags, int lvblen,
+ const struct dlm_lockspace_ops *ops, void *ops_arg,
+ int *ops_result, dlm_lockspace_t **lockspace)
{
struct dlm_ls *ls;
int i, size, error;
int do_unreg = 0;
+ int namelen = strlen(name);
if (namelen > DLM_LOCKSPACE_LEN)
return -EINVAL;
@@ -413,8 +439,24 @@ static int new_lockspace(const char *name, int namelen, void **lockspace,
return -EINVAL;
if (!dlm_user_daemon_available()) {
- module_put(THIS_MODULE);
- return -EUNATCH;
+ log_print("dlm user daemon not available");
+ error = -EUNATCH;
+ goto out;
+ }
+
+ if (ops && ops_result) {
+ if (!dlm_config.ci_recover_callbacks)
+ *ops_result = -EOPNOTSUPP;
+ else
+ *ops_result = 0;
+ }
+
+ if (dlm_config.ci_recover_callbacks && cluster &&
+ strncmp(cluster, dlm_config.ci_cluster_name, DLM_LOCKSPACE_LEN)) {
+ log_print("dlm cluster name %s mismatch %s",
+ dlm_config.ci_cluster_name, cluster);
+ error = -EBADR;
+ goto out;
}
error = 0;
@@ -452,6 +494,11 @@ static int new_lockspace(const char *name, int namelen, void **lockspace,
ls->ls_flags = 0;
ls->ls_scan_time = jiffies;
+ if (ops && dlm_config.ci_recover_callbacks) {
+ ls->ls_ops = ops;
+ ls->ls_ops_arg = ops_arg;
+ }
+
if (flags & DLM_LSFL_TIMEWARN)
set_bit(LSFL_TIMEWARN, &ls->ls_flags);
@@ -463,37 +510,26 @@ static int new_lockspace(const char *name, int namelen, void **lockspace,
size = dlm_config.ci_rsbtbl_size;
ls->ls_rsbtbl_size = size;
- ls->ls_rsbtbl = kmalloc(sizeof(struct dlm_rsbtable) * size, GFP_NOFS);
+ ls->ls_rsbtbl = vmalloc(sizeof(struct dlm_rsbtable) * size);
if (!ls->ls_rsbtbl)
goto out_lsfree;
for (i = 0; i < size; i++) {
- INIT_LIST_HEAD(&ls->ls_rsbtbl[i].list);
- INIT_LIST_HEAD(&ls->ls_rsbtbl[i].toss);
+ ls->ls_rsbtbl[i].keep.rb_node = NULL;
+ ls->ls_rsbtbl[i].toss.rb_node = NULL;
spin_lock_init(&ls->ls_rsbtbl[i].lock);
}
- size = dlm_config.ci_lkbtbl_size;
- ls->ls_lkbtbl_size = size;
+ spin_lock_init(&ls->ls_remove_spin);
- ls->ls_lkbtbl = kmalloc(sizeof(struct dlm_lkbtable) * size, GFP_NOFS);
- if (!ls->ls_lkbtbl)
- goto out_rsbfree;
- for (i = 0; i < size; i++) {
- INIT_LIST_HEAD(&ls->ls_lkbtbl[i].list);
- rwlock_init(&ls->ls_lkbtbl[i].lock);
- ls->ls_lkbtbl[i].counter = 1;
+ for (i = 0; i < DLM_REMOVE_NAMES_MAX; i++) {
+ ls->ls_remove_names[i] = kzalloc(DLM_RESNAME_MAXLEN+1,
+ GFP_KERNEL);
+ if (!ls->ls_remove_names[i])
+ goto out_rsbtbl;
}
- size = dlm_config.ci_dirtbl_size;
- ls->ls_dirtbl_size = size;
-
- ls->ls_dirtbl = kmalloc(sizeof(struct dlm_dirtable) * size, GFP_NOFS);
- if (!ls->ls_dirtbl)
- goto out_lkbfree;
- for (i = 0; i < size; i++) {
- INIT_LIST_HEAD(&ls->ls_dirtbl[i].list);
- spin_lock_init(&ls->ls_dirtbl[i].lock);
- }
+ idr_init(&ls->ls_lkbidr);
+ spin_lock_init(&ls->ls_lkbidr_spin);
INIT_LIST_HEAD(&ls->ls_waiters);
mutex_init(&ls->ls_waiters_mutex);
@@ -502,6 +538,9 @@ static int new_lockspace(const char *name, int namelen, void **lockspace,
INIT_LIST_HEAD(&ls->ls_timeout);
mutex_init(&ls->ls_timeout_mutex);
+ INIT_LIST_HEAD(&ls->ls_new_rsb);
+ spin_lock_init(&ls->ls_new_rsb_spin);
+
INIT_LIST_HEAD(&ls->ls_nodes);
INIT_LIST_HEAD(&ls->ls_nodes_gone);
ls->ls_num_nodes = 0;
@@ -520,6 +559,9 @@ static int new_lockspace(const char *name, int namelen, void **lockspace,
init_completion(&ls->ls_members_done);
ls->ls_members_result = -1;
+ mutex_init(&ls->ls_cb_mutex);
+ INIT_LIST_HEAD(&ls->ls_cb_delay);
+
ls->ls_recoverd_task = NULL;
mutex_init(&ls->ls_recoverd_active);
spin_lock_init(&ls->ls_recover_lock);
@@ -536,35 +578,59 @@ static int new_lockspace(const char *name, int namelen, void **lockspace,
ls->ls_recover_buf = kmalloc(dlm_config.ci_buffer_size, GFP_NOFS);
if (!ls->ls_recover_buf)
- goto out_dirfree;
+ goto out_lkbidr;
+
+ ls->ls_slot = 0;
+ ls->ls_num_slots = 0;
+ ls->ls_slots_size = 0;
+ ls->ls_slots = NULL;
INIT_LIST_HEAD(&ls->ls_recover_list);
spin_lock_init(&ls->ls_recover_list_lock);
+ idr_init(&ls->ls_recover_idr);
+ spin_lock_init(&ls->ls_recover_idr_lock);
ls->ls_recover_list_count = 0;
ls->ls_local_handle = ls;
init_waitqueue_head(&ls->ls_wait_general);
INIT_LIST_HEAD(&ls->ls_root_list);
init_rwsem(&ls->ls_root_sem);
- down_write(&ls->ls_in_recovery);
-
spin_lock(&lslist_lock);
ls->ls_create_count = 1;
list_add(&ls->ls_list, &lslist);
spin_unlock(&lslist_lock);
- /* needs to find ls in lslist */
+ if (flags & DLM_LSFL_FS) {
+ error = dlm_callback_start(ls);
+ if (error) {
+ log_error(ls, "can't start dlm_callback %d", error);
+ goto out_delist;
+ }
+ }
+
+ init_waitqueue_head(&ls->ls_recover_lock_wait);
+
+ /*
+ * Once started, dlm_recoverd first looks for ls in lslist, then
+ * initializes ls_in_recovery as locked in "down" mode. We need
+ * to wait for the wakeup from dlm_recoverd because in_recovery
+ * has to start out in down mode.
+ */
+
error = dlm_recoverd_start(ls);
if (error) {
log_error(ls, "can't start dlm_recoverd %d", error);
- goto out_delist;
+ goto out_callback;
}
+ wait_event(ls->ls_recover_lock_wait,
+ test_bit(LSFL_RECOVER_LOCK, &ls->ls_flags));
+
ls->ls_kobj.kset = dlm_kset;
error = kobject_init_and_add(&ls->ls_kobj, &dlm_ktype, NULL,
"%s", ls->ls_name);
if (error)
- goto out_stop;
+ goto out_recoverd;
kobject_uevent(&ls->ls_kobj, KOBJ_ADD);
/* let kobject handle freeing of ls if there's an error */
@@ -578,7 +644,7 @@ static int new_lockspace(const char *name, int namelen, void **lockspace,
error = do_uevent(ls, 1);
if (error)
- goto out_stop;
+ goto out_recoverd;
wait_for_completion(&ls->ls_members_done);
error = ls->ls_members_result;
@@ -587,7 +653,7 @@ static int new_lockspace(const char *name, int namelen, void **lockspace,
dlm_create_debug_file(ls);
- log_debug(ls, "join complete");
+ log_rinfo(ls, "join complete");
*lockspace = ls;
return 0;
@@ -595,19 +661,24 @@ static int new_lockspace(const char *name, int namelen, void **lockspace,
do_uevent(ls, 0);
dlm_clear_members(ls);
kfree(ls->ls_node_array);
- out_stop:
+ out_recoverd:
dlm_recoverd_stop(ls);
+ out_callback:
+ dlm_callback_stop(ls);
out_delist:
spin_lock(&lslist_lock);
list_del(&ls->ls_list);
spin_unlock(&lslist_lock);
+ idr_destroy(&ls->ls_recover_idr);
kfree(ls->ls_recover_buf);
- out_dirfree:
- kfree(ls->ls_dirtbl);
- out_lkbfree:
- kfree(ls->ls_lkbtbl);
- out_rsbfree:
- kfree(ls->ls_rsbtbl);
+ out_lkbidr:
+ idr_destroy(&ls->ls_lkbidr);
+ for (i = 0; i < DLM_REMOVE_NAMES_MAX; i++) {
+ if (ls->ls_remove_names[i])
+ kfree(ls->ls_remove_names[i]);
+ }
+ out_rsbtbl:
+ vfree(ls->ls_rsbtbl);
out_lsfree:
if (do_unreg)
kobject_put(&ls->ls_kobj);
@@ -618,8 +689,10 @@ static int new_lockspace(const char *name, int namelen, void **lockspace,
return error;
}
-int dlm_new_lockspace(const char *name, int namelen, void **lockspace,
- uint32_t flags, int lvblen)
+int dlm_new_lockspace(const char *name, const char *cluster,
+ uint32_t flags, int lvblen,
+ const struct dlm_lockspace_ops *ops, void *ops_arg,
+ int *ops_result, dlm_lockspace_t **lockspace)
{
int error = 0;
@@ -629,7 +702,8 @@ int dlm_new_lockspace(const char *name, int namelen, void **lockspace,
if (error)
goto out;
- error = new_lockspace(name, namelen, lockspace, flags, lvblen);
+ error = new_lockspace(name, cluster, flags, lvblen, ops, ops_arg,
+ ops_result, lockspace);
if (!error)
ls_count++;
if (error > 0)
@@ -641,50 +715,62 @@ int dlm_new_lockspace(const char *name, int namelen, void **lockspace,
return error;
}
-/* Return 1 if the lockspace still has active remote locks,
- * 2 if the lockspace still has active local locks.
- */
-static int lockspace_busy(struct dlm_ls *ls)
-{
- int i, lkb_found = 0;
- struct dlm_lkb *lkb;
-
- /* NOTE: We check the lockidtbl here rather than the resource table.
- This is because there may be LKBs queued as ASTs that have been
- unlinked from their RSBs and are pending deletion once the AST has
- been delivered */
-
- for (i = 0; i < ls->ls_lkbtbl_size; i++) {
- read_lock(&ls->ls_lkbtbl[i].lock);
- if (!list_empty(&ls->ls_lkbtbl[i].list)) {
- lkb_found = 1;
- list_for_each_entry(lkb, &ls->ls_lkbtbl[i].list,
- lkb_idtbl_list) {
- if (!lkb->lkb_nodeid) {
- read_unlock(&ls->ls_lkbtbl[i].lock);
- return 2;
- }
- }
- }
- read_unlock(&ls->ls_lkbtbl[i].lock);
+static int lkb_idr_is_local(int id, void *p, void *data)
+{
+ struct dlm_lkb *lkb = p;
+
+ return lkb->lkb_nodeid == 0 && lkb->lkb_grmode != DLM_LOCK_IV;
+}
+
+static int lkb_idr_is_any(int id, void *p, void *data)
+{
+ return 1;
+}
+
+static int lkb_idr_free(int id, void *p, void *data)
+{
+ struct dlm_lkb *lkb = p;
+
+ if (lkb->lkb_lvbptr && lkb->lkb_flags & DLM_IFL_MSTCPY)
+ dlm_free_lvb(lkb->lkb_lvbptr);
+
+ dlm_free_lkb(lkb);
+ return 0;
+}
+
+/* NOTE: We check the lkbidr here rather than the resource table.
+ This is because there may be LKBs queued as ASTs that have been unlinked
+ from their RSBs and are pending deletion once the AST has been delivered */
+
+static int lockspace_busy(struct dlm_ls *ls, int force)
+{
+ int rv;
+
+ spin_lock(&ls->ls_lkbidr_spin);
+ if (force == 0) {
+ rv = idr_for_each(&ls->ls_lkbidr, lkb_idr_is_any, ls);
+ } else if (force == 1) {
+ rv = idr_for_each(&ls->ls_lkbidr, lkb_idr_is_local, ls);
+ } else {
+ rv = 0;
}
- return lkb_found;
+ spin_unlock(&ls->ls_lkbidr_spin);
+ return rv;
}
static int release_lockspace(struct dlm_ls *ls, int force)
{
- struct dlm_lkb *lkb;
struct dlm_rsb *rsb;
- struct list_head *head;
+ struct rb_node *n;
int i, busy, rv;
- busy = lockspace_busy(ls);
+ busy = lockspace_busy(ls, force);
spin_lock(&lslist_lock);
if (ls->ls_create_count == 1) {
- if (busy > force)
+ if (busy) {
rv = -EBUSY;
- else {
+ } else {
/* remove_lockspace takes ls off lslist */
ls->ls_create_count = 0;
rv = 0;
@@ -708,69 +794,50 @@ static int release_lockspace(struct dlm_ls *ls, int force)
dlm_recoverd_stop(ls);
+ dlm_callback_stop(ls);
+
remove_lockspace(ls);
dlm_delete_debug_file(ls);
- dlm_astd_suspend();
-
kfree(ls->ls_recover_buf);
/*
- * Free direntry structs.
+ * Free all lkb's in idr
*/
- dlm_dir_clear(ls);
- kfree(ls->ls_dirtbl);
-
- /*
- * Free all lkb's on lkbtbl[] lists.
- */
-
- for (i = 0; i < ls->ls_lkbtbl_size; i++) {
- head = &ls->ls_lkbtbl[i].list;
- while (!list_empty(head)) {
- lkb = list_entry(head->next, struct dlm_lkb,
- lkb_idtbl_list);
-
- list_del(&lkb->lkb_idtbl_list);
-
- dlm_del_ast(lkb);
-
- if (lkb->lkb_lvbptr && lkb->lkb_flags & DLM_IFL_MSTCPY)
- dlm_free_lvb(lkb->lkb_lvbptr);
-
- dlm_free_lkb(lkb);
- }
- }
- dlm_astd_resume();
-
- kfree(ls->ls_lkbtbl);
+ idr_for_each(&ls->ls_lkbidr, lkb_idr_free, ls);
+ idr_destroy(&ls->ls_lkbidr);
/*
* Free all rsb's on rsbtbl[] lists
*/
for (i = 0; i < ls->ls_rsbtbl_size; i++) {
- head = &ls->ls_rsbtbl[i].list;
- while (!list_empty(head)) {
- rsb = list_entry(head->next, struct dlm_rsb,
- res_hashchain);
-
- list_del(&rsb->res_hashchain);
+ while ((n = rb_first(&ls->ls_rsbtbl[i].keep))) {
+ rsb = rb_entry(n, struct dlm_rsb, res_hashnode);
+ rb_erase(n, &ls->ls_rsbtbl[i].keep);
dlm_free_rsb(rsb);
}
- head = &ls->ls_rsbtbl[i].toss;
- while (!list_empty(head)) {
- rsb = list_entry(head->next, struct dlm_rsb,
- res_hashchain);
- list_del(&rsb->res_hashchain);
+ while ((n = rb_first(&ls->ls_rsbtbl[i].toss))) {
+ rsb = rb_entry(n, struct dlm_rsb, res_hashnode);
+ rb_erase(n, &ls->ls_rsbtbl[i].toss);
dlm_free_rsb(rsb);
}
}
- kfree(ls->ls_rsbtbl);
+ vfree(ls->ls_rsbtbl);
+
+ for (i = 0; i < DLM_REMOVE_NAMES_MAX; i++)
+ kfree(ls->ls_remove_names[i]);
+
+ while (!list_empty(&ls->ls_new_rsb)) {
+ rsb = list_first_entry(&ls->ls_new_rsb, struct dlm_rsb,
+ res_hashchain);
+ list_del(&rsb->res_hashchain);
+ dlm_free_rsb(rsb);
+ }
/*
* Free structures on any other lists
@@ -778,11 +845,10 @@ static int release_lockspace(struct dlm_ls *ls, int force)
dlm_purge_requestqueue(ls);
kfree(ls->ls_recover_args);
- dlm_clear_free_entries(ls);
dlm_clear_members(ls);
dlm_clear_members_gone(ls);
kfree(ls->ls_node_array);
- log_debug(ls, "release_lockspace final free");
+ log_rinfo(ls, "release_lockspace final free");
kobject_put(&ls->ls_kobj);
/* The ls structure will be freed when the kobject is done with */
@@ -828,17 +894,24 @@ int dlm_release_lockspace(void *lockspace, int force)
void dlm_stop_lockspaces(void)
{
struct dlm_ls *ls;
+ int count;
restart:
+ count = 0;
spin_lock(&lslist_lock);
list_for_each_entry(ls, &lslist, ls_list) {
- if (!test_bit(LSFL_RUNNING, &ls->ls_flags))
+ if (!test_bit(LSFL_RUNNING, &ls->ls_flags)) {
+ count++;
continue;
+ }
spin_unlock(&lslist_lock);
log_error(ls, "no userland control daemon, stopping lockspace");
dlm_ls_stop(ls);
goto restart;
}
spin_unlock(&lslist_lock);
+
+ if (count)
+ log_print("dlm user daemon left %d lockspaces", count);
}
diff --git a/fs/dlm/lowcomms.c b/fs/dlm/lowcomms.c
index 37a34c2c622..d08e079ea5d 100644
--- a/fs/dlm/lowcomms.c
+++ b/fs/dlm/lowcomms.c
@@ -52,7 +52,7 @@
#include <linux/mutex.h>
#include <linux/sctp.h>
#include <linux/slab.h>
-#include <net/sctp/user.h>
+#include <net/sctp/sctp.h>
#include <net/ipv6.h>
#include "dlm_internal.h"
@@ -63,6 +63,9 @@
#define NEEDED_RMEM (4*1024*1024)
#define CONN_HASH_SIZE 32
+/* Number of messages to send before rescheduling */
+#define MAX_SEND_MSG_COUNT 25
+
struct cbuf {
unsigned int base;
unsigned int len;
@@ -108,6 +111,7 @@ struct connection {
#define CF_INIT_PENDING 4
#define CF_IS_OTHERCON 5
#define CF_CLOSE 6
+#define CF_APP_LIMITED 7
struct list_head writequeue; /* List of outgoing writequeue_entries */
spinlock_t writequeue_lock;
int (*rx_action) (struct connection *); /* What to do when active */
@@ -121,6 +125,7 @@ struct connection {
struct connection *othercon;
struct work_struct rwork; /* Receive workqueue */
struct work_struct swork; /* Send workqueue */
+ bool try_new_addr;
};
#define sock2con(x) ((struct connection *)(x)->sk_user_data)
@@ -135,8 +140,20 @@ struct writequeue_entry {
struct connection *con;
};
+struct dlm_node_addr {
+ struct list_head list;
+ int nodeid;
+ int addr_count;
+ int curr_addr_index;
+ struct sockaddr_storage *addr[DLM_MAX_ADDR_COUNT];
+};
+
+static LIST_HEAD(dlm_node_addrs);
+static DEFINE_SPINLOCK(dlm_node_addrs_spin);
+
static struct sockaddr_storage *dlm_local_addr[DLM_MAX_ADDR_COUNT];
static int dlm_local_count;
+static int dlm_allow_conn;
/* Work queues */
static struct workqueue_struct *recv_workqueue;
@@ -161,12 +178,11 @@ static inline int nodeid_hash(int nodeid)
static struct connection *__find_con(int nodeid)
{
int r;
- struct hlist_node *h;
struct connection *con;
r = nodeid_hash(nodeid);
- hlist_for_each_entry(con, h, &connection_hash[r], list) {
+ hlist_for_each_entry(con, &connection_hash[r], list) {
if (con->nodeid == nodeid)
return con;
}
@@ -216,13 +232,12 @@ static struct connection *__nodeid2con(int nodeid, gfp_t alloc)
static void foreach_conn(void (*conn_func)(struct connection *c))
{
int i;
- struct hlist_node *h, *n;
+ struct hlist_node *n;
struct connection *con;
for (i = 0; i < CONN_HASH_SIZE; i++) {
- hlist_for_each_entry_safe(con, h, n, &connection_hash[i], list){
+ hlist_for_each_entry_safe(con, n, &connection_hash[i], list)
conn_func(con);
- }
}
}
@@ -241,13 +256,12 @@ static struct connection *nodeid2con(int nodeid, gfp_t allocation)
static struct connection *assoc2con(int assoc_id)
{
int i;
- struct hlist_node *h;
struct connection *con;
mutex_lock(&connections_lock);
for (i = 0 ; i < CONN_HASH_SIZE; i++) {
- hlist_for_each_entry(con, h, &connection_hash[i], list) {
+ hlist_for_each_entry(con, &connection_hash[i], list) {
if (con->sctp_assoc == assoc_id) {
mutex_unlock(&connections_lock);
return con;
@@ -258,33 +272,159 @@ static struct connection *assoc2con(int assoc_id)
return NULL;
}
-static int nodeid_to_addr(int nodeid, struct sockaddr *retaddr)
+static struct dlm_node_addr *find_node_addr(int nodeid)
+{
+ struct dlm_node_addr *na;
+
+ list_for_each_entry(na, &dlm_node_addrs, list) {
+ if (na->nodeid == nodeid)
+ return na;
+ }
+ return NULL;
+}
+
+static int addr_compare(struct sockaddr_storage *x, struct sockaddr_storage *y)
+{
+ switch (x->ss_family) {
+ case AF_INET: {
+ struct sockaddr_in *sinx = (struct sockaddr_in *)x;
+ struct sockaddr_in *siny = (struct sockaddr_in *)y;
+ if (sinx->sin_addr.s_addr != siny->sin_addr.s_addr)
+ return 0;
+ if (sinx->sin_port != siny->sin_port)
+ return 0;
+ break;
+ }
+ case AF_INET6: {
+ struct sockaddr_in6 *sinx = (struct sockaddr_in6 *)x;
+ struct sockaddr_in6 *siny = (struct sockaddr_in6 *)y;
+ if (!ipv6_addr_equal(&sinx->sin6_addr, &siny->sin6_addr))
+ return 0;
+ if (sinx->sin6_port != siny->sin6_port)
+ return 0;
+ break;
+ }
+ default:
+ return 0;
+ }
+ return 1;
+}
+
+static int nodeid_to_addr(int nodeid, struct sockaddr_storage *sas_out,
+ struct sockaddr *sa_out, bool try_new_addr)
{
- struct sockaddr_storage addr;
- int error;
+ struct sockaddr_storage sas;
+ struct dlm_node_addr *na;
if (!dlm_local_count)
return -1;
- error = dlm_nodeid_to_addr(nodeid, &addr);
- if (error)
- return error;
+ spin_lock(&dlm_node_addrs_spin);
+ na = find_node_addr(nodeid);
+ if (na && na->addr_count) {
+ if (try_new_addr) {
+ na->curr_addr_index++;
+ if (na->curr_addr_index == na->addr_count)
+ na->curr_addr_index = 0;
+ }
+
+ memcpy(&sas, na->addr[na->curr_addr_index ],
+ sizeof(struct sockaddr_storage));
+ }
+ spin_unlock(&dlm_node_addrs_spin);
+
+ if (!na)
+ return -EEXIST;
+
+ if (!na->addr_count)
+ return -ENOENT;
+
+ if (sas_out)
+ memcpy(sas_out, &sas, sizeof(struct sockaddr_storage));
+
+ if (!sa_out)
+ return 0;
if (dlm_local_addr[0]->ss_family == AF_INET) {
- struct sockaddr_in *in4 = (struct sockaddr_in *) &addr;
- struct sockaddr_in *ret4 = (struct sockaddr_in *) retaddr;
+ struct sockaddr_in *in4 = (struct sockaddr_in *) &sas;
+ struct sockaddr_in *ret4 = (struct sockaddr_in *) sa_out;
ret4->sin_addr.s_addr = in4->sin_addr.s_addr;
} else {
- struct sockaddr_in6 *in6 = (struct sockaddr_in6 *) &addr;
- struct sockaddr_in6 *ret6 = (struct sockaddr_in6 *) retaddr;
- ipv6_addr_copy(&ret6->sin6_addr, &in6->sin6_addr);
+ struct sockaddr_in6 *in6 = (struct sockaddr_in6 *) &sas;
+ struct sockaddr_in6 *ret6 = (struct sockaddr_in6 *) sa_out;
+ ret6->sin6_addr = in6->sin6_addr;
+ }
+
+ return 0;
+}
+
+static int addr_to_nodeid(struct sockaddr_storage *addr, int *nodeid)
+{
+ struct dlm_node_addr *na;
+ int rv = -EEXIST;
+ int addr_i;
+
+ spin_lock(&dlm_node_addrs_spin);
+ list_for_each_entry(na, &dlm_node_addrs, list) {
+ if (!na->addr_count)
+ continue;
+
+ for (addr_i = 0; addr_i < na->addr_count; addr_i++) {
+ if (addr_compare(na->addr[addr_i], addr)) {
+ *nodeid = na->nodeid;
+ rv = 0;
+ goto unlock;
+ }
+ }
+ }
+unlock:
+ spin_unlock(&dlm_node_addrs_spin);
+ return rv;
+}
+
+int dlm_lowcomms_addr(int nodeid, struct sockaddr_storage *addr, int len)
+{
+ struct sockaddr_storage *new_addr;
+ struct dlm_node_addr *new_node, *na;
+
+ new_node = kzalloc(sizeof(struct dlm_node_addr), GFP_NOFS);
+ if (!new_node)
+ return -ENOMEM;
+
+ new_addr = kzalloc(sizeof(struct sockaddr_storage), GFP_NOFS);
+ if (!new_addr) {
+ kfree(new_node);
+ return -ENOMEM;
+ }
+
+ memcpy(new_addr, addr, len);
+
+ spin_lock(&dlm_node_addrs_spin);
+ na = find_node_addr(nodeid);
+ if (!na) {
+ new_node->nodeid = nodeid;
+ new_node->addr[0] = new_addr;
+ new_node->addr_count = 1;
+ list_add(&new_node->list, &dlm_node_addrs);
+ spin_unlock(&dlm_node_addrs_spin);
+ return 0;
+ }
+
+ if (na->addr_count >= DLM_MAX_ADDR_COUNT) {
+ spin_unlock(&dlm_node_addrs_spin);
+ kfree(new_addr);
+ kfree(new_node);
+ return -ENOSPC;
}
+ na->addr[na->addr_count++] = new_addr;
+ spin_unlock(&dlm_node_addrs_spin);
+ kfree(new_node);
return 0;
}
/* Data available on socket or listen socket received a connect */
-static void lowcomms_data_ready(struct sock *sk, int count_unused)
+static void lowcomms_data_ready(struct sock *sk)
{
struct connection *con = sock2con(sk);
if (con && !test_and_set_bit(CF_READ_PENDING, &con->flags))
@@ -295,7 +435,17 @@ static void lowcomms_write_space(struct sock *sk)
{
struct connection *con = sock2con(sk);
- if (con && !test_and_set_bit(CF_WRITE_PENDING, &con->flags))
+ if (!con)
+ return;
+
+ clear_bit(SOCK_NOSPACE, &con->sock->flags);
+
+ if (test_and_clear_bit(CF_APP_LIMITED, &con->flags)) {
+ con->sock->sk->sk_write_pending--;
+ clear_bit(SOCK_ASYNC_NOSPACE, &con->sock->flags);
+ }
+
+ if (!test_and_set_bit(CF_WRITE_PENDING, &con->flags))
queue_work(send_workqueue, &con->swork);
}
@@ -332,7 +482,7 @@ int dlm_lowcomms_connect_node(int nodeid)
}
/* Make a socket active */
-static int add_sock(struct socket *sock, struct connection *con)
+static void add_sock(struct socket *sock, struct connection *con)
{
con->sock = sock;
@@ -342,7 +492,6 @@ static int add_sock(struct socket *sock, struct connection *con)
con->sock->sk->sk_state_change = lowcomms_state_change;
con->sock->sk->sk_user_data = con;
con->sock->sk->sk_allocation = GFP_NOFS;
- return 0;
}
/* Add the port number to an IPv6 or 4 sockaddr and return the address
@@ -424,8 +573,23 @@ static void sctp_send_shutdown(sctp_assoc_t associd)
static void sctp_init_failed_foreach(struct connection *con)
{
+
+ /*
+ * Don't try to recover base con and handle race where the
+ * other node's assoc init creates a assoc and we get that
+ * notification, then we get a notification that our attempt
+ * failed due. This happens when we are still trying the primary
+ * address, but the other node has already tried secondary addrs
+ * and found one that worked.
+ */
+ if (!con->nodeid || con->sctp_assoc)
+ return;
+
+ log_print("Retrying SCTP association init for node %d\n", con->nodeid);
+
+ con->try_new_addr = true;
con->sctp_assoc = 0;
- if (test_and_clear_bit(CF_CONNECT_PENDING, &con->flags)) {
+ if (test_and_clear_bit(CF_INIT_PENDING, &con->flags)) {
if (!test_and_set_bit(CF_WRITE_PENDING, &con->flags))
queue_work(send_workqueue, &con->swork);
}
@@ -442,15 +606,62 @@ static void sctp_init_failed(void)
mutex_unlock(&connections_lock);
}
+static void retry_failed_sctp_send(struct connection *recv_con,
+ struct sctp_send_failed *sn_send_failed,
+ char *buf)
+{
+ int len = sn_send_failed->ssf_length - sizeof(struct sctp_send_failed);
+ struct dlm_mhandle *mh;
+ struct connection *con;
+ char *retry_buf;
+ int nodeid = sn_send_failed->ssf_info.sinfo_ppid;
+
+ log_print("Retry sending %d bytes to node id %d", len, nodeid);
+
+ if (!nodeid) {
+ log_print("Shouldn't resend data via listening connection.");
+ return;
+ }
+
+ con = nodeid2con(nodeid, 0);
+ if (!con) {
+ log_print("Could not look up con for nodeid %d\n",
+ nodeid);
+ return;
+ }
+
+ mh = dlm_lowcomms_get_buffer(nodeid, len, GFP_NOFS, &retry_buf);
+ if (!mh) {
+ log_print("Could not allocate buf for retry.");
+ return;
+ }
+ memcpy(retry_buf, buf + sizeof(struct sctp_send_failed), len);
+ dlm_lowcomms_commit_buffer(mh);
+
+ /*
+ * If we got a assoc changed event before the send failed event then
+ * we only need to retry the send.
+ */
+ if (con->sctp_assoc) {
+ if (!test_and_set_bit(CF_WRITE_PENDING, &con->flags))
+ queue_work(send_workqueue, &con->swork);
+ } else
+ sctp_init_failed_foreach(con);
+}
+
/* Something happened to an association */
static void process_sctp_notification(struct connection *con,
struct msghdr *msg, char *buf)
{
union sctp_notification *sn = (union sctp_notification *)buf;
+ struct linger linger;
- if (sn->sn_header.sn_type == SCTP_ASSOC_CHANGE) {
+ switch (sn->sn_header.sn_type) {
+ case SCTP_SEND_FAILED:
+ retry_failed_sctp_send(con, &sn->sn_send_failed, buf);
+ break;
+ case SCTP_ASSOC_CHANGE:
switch (sn->sn_assoc_change.sac_state) {
-
case SCTP_COMM_UP:
case SCTP_RESTART:
{
@@ -460,9 +671,6 @@ static void process_sctp_notification(struct connection *con,
int prim_len, ret;
int addr_len;
struct connection *new_con;
- sctp_peeloff_arg_t parg;
- int parglen = sizeof(parg);
- int err;
/*
* We get this before any data for an association.
@@ -497,13 +705,11 @@ static void process_sctp_notification(struct connection *con,
return;
}
make_sockaddr(&prim.ssp_addr, 0, &addr_len);
- if (dlm_addr_to_nodeid(&prim.ssp_addr, &nodeid)) {
- int i;
+ if (addr_to_nodeid(&prim.ssp_addr, &nodeid)) {
unsigned char *b=(unsigned char *)&prim.ssp_addr;
log_print("reject connect from unknown addr");
- for (i=0; i<sizeof(struct sockaddr_storage);i++)
- printk("%02x ", b[i]);
- printk("\n");
+ print_hex_dump_bytes("ss: ", DUMP_PREFIX_NONE,
+ b, sizeof(struct sockaddr_storage));
sctp_send_shutdown(prim.ssp_assoc_id);
return;
}
@@ -513,30 +719,35 @@ static void process_sctp_notification(struct connection *con,
return;
/* Peel off a new sock */
- parg.associd = sn->sn_assoc_change.sac_assoc_id;
- ret = kernel_getsockopt(con->sock, IPPROTO_SCTP,
- SCTP_SOCKOPT_PEELOFF,
- (void *)&parg, &parglen);
+ lock_sock(con->sock->sk);
+ ret = sctp_do_peeloff(con->sock->sk,
+ sn->sn_assoc_change.sac_assoc_id,
+ &new_con->sock);
+ release_sock(con->sock->sk);
if (ret < 0) {
log_print("Can't peel off a socket for "
"connection %d to node %d: err=%d",
- parg.associd, nodeid, ret);
- return;
- }
- new_con->sock = sockfd_lookup(parg.sd, &err);
- if (!new_con->sock) {
- log_print("sockfd_lookup error %d", err);
+ (int)sn->sn_assoc_change.sac_assoc_id,
+ nodeid, ret);
return;
}
add_sock(new_con->sock, new_con);
- sockfd_put(new_con->sock);
+
+ linger.l_onoff = 1;
+ linger.l_linger = 0;
+ ret = kernel_setsockopt(new_con->sock, SOL_SOCKET, SO_LINGER,
+ (char *)&linger, sizeof(linger));
+ if (ret < 0)
+ log_print("set socket option SO_LINGER failed");
log_print("connecting to %d sctp association %d",
nodeid, (int)sn->sn_assoc_change.sac_assoc_id);
+ new_con->sctp_assoc = sn->sn_assoc_change.sac_assoc_id;
+ new_con->try_new_addr = false;
/* Send any pending writes */
clear_bit(CF_CONNECT_PENDING, &new_con->flags);
- clear_bit(CF_INIT_PENDING, &con->flags);
+ clear_bit(CF_INIT_PENDING, &new_con->flags);
if (!test_and_set_bit(CF_WRITE_PENDING, &new_con->flags)) {
queue_work(send_workqueue, &new_con->swork);
}
@@ -555,14 +766,10 @@ static void process_sctp_notification(struct connection *con,
}
break;
- /* We don't know which INIT failed, so clear the PENDING flags
- * on them all. if assoc_id is zero then it will then try
- * again */
-
case SCTP_CANT_STR_ASSOC:
{
+ /* Will retry init when we get the send failed notification */
log_print("Can't start SCTP association - retrying");
- sctp_init_failed();
}
break;
@@ -571,6 +778,8 @@ static void process_sctp_notification(struct connection *con,
(int)sn->sn_assoc_change.sac_assoc_id,
sn->sn_assoc_change.sac_state);
}
+ default:
+ ; /* fall through */
}
}
@@ -704,6 +913,13 @@ static int tcp_accept_from_sock(struct connection *con)
struct connection *newcon;
struct connection *addcon;
+ mutex_lock(&connections_lock);
+ if (!dlm_allow_conn) {
+ mutex_unlock(&connections_lock);
+ return -1;
+ }
+ mutex_unlock(&connections_lock);
+
memset(&peeraddr, 0, sizeof(peeraddr));
result = sock_create_kern(dlm_local_addr[0]->ss_family, SOCK_STREAM,
IPPROTO_TCP, &newsock);
@@ -733,8 +949,11 @@ static int tcp_accept_from_sock(struct connection *con)
/* Get the new node's NODEID */
make_sockaddr(&peeraddr, 0, &len);
- if (dlm_addr_to_nodeid(&peeraddr, &nodeid)) {
+ if (addr_to_nodeid(&peeraddr, &nodeid)) {
+ unsigned char *b=(unsigned char *)&peeraddr;
log_print("connect from non cluster node");
+ print_hex_dump_bytes("ss: ", DUMP_PREFIX_NONE,
+ b, sizeof(struct sockaddr_storage));
sock_release(newsock);
mutex_unlock(&con->sock_mutex);
return -1;
@@ -796,7 +1015,7 @@ static int tcp_accept_from_sock(struct connection *con)
/*
* Add it to the active queue in case we got data
- * beween processing the accept adding the socket
+ * between processing the accept adding the socket
* to the read_sockets list
*/
if (!test_and_set_bit(CF_READ_PENDING, &addcon->flags))
@@ -820,6 +1039,24 @@ static void free_entry(struct writequeue_entry *e)
kfree(e);
}
+/*
+ * writequeue_entry_complete - try to delete and free write queue entry
+ * @e: write queue entry to try to delete
+ * @completed: bytes completed
+ *
+ * writequeue_lock must be held.
+ */
+static void writequeue_entry_complete(struct writequeue_entry *e, int completed)
+{
+ e->offset += completed;
+ e->len -= completed;
+
+ if (e->len == 0 && e->users == 0) {
+ list_del(&e->list);
+ free_entry(e);
+ }
+}
+
/* Initiate an SCTP association.
This is a special case of send_to_sock() in that we don't yet have a
peeled-off socket for this association, so we use the listening socket
@@ -839,15 +1076,14 @@ static void sctp_init_assoc(struct connection *con)
int addrlen;
struct kvec iov[1];
+ mutex_lock(&con->sock_mutex);
if (test_and_set_bit(CF_INIT_PENDING, &con->flags))
- return;
-
- if (con->retries++ > MAX_CONNECT_RETRIES)
- return;
+ goto unlock;
- if (nodeid_to_addr(con->nodeid, (struct sockaddr *)&rem_addr)) {
+ if (nodeid_to_addr(con->nodeid, NULL, (struct sockaddr *)&rem_addr,
+ con->try_new_addr)) {
log_print("no address for nodeid %d", con->nodeid);
- return;
+ goto unlock;
}
base_con = nodeid2con(0, 0);
BUG_ON(base_con == NULL);
@@ -865,17 +1101,25 @@ static void sctp_init_assoc(struct connection *con)
if (list_empty(&con->writequeue)) {
spin_unlock(&con->writequeue_lock);
log_print("writequeue empty for nodeid %d", con->nodeid);
- return;
+ goto unlock;
}
e = list_first_entry(&con->writequeue, struct writequeue_entry, list);
len = e->len;
offset = e->offset;
- spin_unlock(&con->writequeue_lock);
/* Send the first block off the write queue */
iov[0].iov_base = page_address(e->page)+offset;
iov[0].iov_len = len;
+ spin_unlock(&con->writequeue_lock);
+
+ if (rem_addr.ss_family == AF_INET) {
+ struct sockaddr_in *sin = (struct sockaddr_in *)&rem_addr;
+ log_print("Trying to connect to %pI4", &sin->sin_addr.s_addr);
+ } else {
+ struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)&rem_addr;
+ log_print("Trying to connect to %pI6", &sin6->sin6_addr);
+ }
cmsg = CMSG_FIRSTHDR(&outmessage);
cmsg->cmsg_level = IPPROTO_SCTP;
@@ -883,8 +1127,9 @@ static void sctp_init_assoc(struct connection *con)
cmsg->cmsg_len = CMSG_LEN(sizeof(struct sctp_sndrcvinfo));
sinfo = CMSG_DATA(cmsg);
memset(sinfo, 0x00, sizeof(struct sctp_sndrcvinfo));
- sinfo->sinfo_ppid = cpu_to_le32(dlm_our_nodeid());
+ sinfo->sinfo_ppid = cpu_to_le32(con->nodeid);
outmessage.msg_controllen = cmsg->cmsg_len;
+ sinfo->sinfo_flags |= SCTP_ADDR_OVER;
ret = kernel_sendmsg(base_con->sock, &outmessage, iov, 1, len);
if (ret < 0) {
@@ -897,24 +1142,22 @@ static void sctp_init_assoc(struct connection *con)
}
else {
spin_lock(&con->writequeue_lock);
- e->offset += ret;
- e->len -= ret;
-
- if (e->len == 0 && e->users == 0) {
- list_del(&e->list);
- free_entry(e);
- }
+ writequeue_entry_complete(e, ret);
spin_unlock(&con->writequeue_lock);
}
+
+unlock:
+ mutex_unlock(&con->sock_mutex);
}
/* Connect a new socket to its peer */
static void tcp_connect_to_sock(struct connection *con)
{
- int result = -EHOSTUNREACH;
struct sockaddr_storage saddr, src_addr;
int addr_len;
struct socket *sock = NULL;
+ int one = 1;
+ int result;
if (con->nodeid == 0) {
log_print("attempt to connect sock 0 foiled");
@@ -926,10 +1169,8 @@ static void tcp_connect_to_sock(struct connection *con)
goto out;
/* Some odd races can cause double-connects, ignore them */
- if (con->sock) {
- result = 0;
+ if (con->sock)
goto out;
- }
/* Create a socket to communicate with */
result = sock_create_kern(dlm_local_addr[0]->ss_family, SOCK_STREAM,
@@ -938,8 +1179,11 @@ static void tcp_connect_to_sock(struct connection *con)
goto out_err;
memset(&saddr, 0, sizeof(saddr));
- if (dlm_nodeid_to_addr(con->nodeid, &saddr))
+ result = nodeid_to_addr(con->nodeid, &saddr, NULL, false);
+ if (result < 0) {
+ log_print("no address for nodeid %d", con->nodeid);
goto out_err;
+ }
sock->sk->sk_user_data = con;
con->rx_action = receive_from_sock;
@@ -960,8 +1204,12 @@ static void tcp_connect_to_sock(struct connection *con)
make_sockaddr(&saddr, dlm_config.ci_tcp_port, &addr_len);
log_print("connecting to %d", con->nodeid);
- result =
- sock->ops->connect(sock, (struct sockaddr *)&saddr, addr_len,
+
+ /* Turn off Nagle's algorithm */
+ kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY, (char *)&one,
+ sizeof(one));
+
+ result = sock->ops->connect(sock, (struct sockaddr *)&saddr, addr_len,
O_NONBLOCK);
if (result == -EINPROGRESS)
result = 0;
@@ -979,11 +1227,17 @@ out_err:
* Some errors are fatal and this list might need adjusting. For other
* errors we try again until the max number of retries is reached.
*/
- if (result != -EHOSTUNREACH && result != -ENETUNREACH &&
- result != -ENETDOWN && result != -EINVAL
- && result != -EPROTONOSUPPORT) {
+ if (result != -EHOSTUNREACH &&
+ result != -ENETUNREACH &&
+ result != -ENETDOWN &&
+ result != -EINVAL &&
+ result != -EPROTONOSUPPORT) {
+ log_print("connect %d try %d error %d", con->nodeid,
+ con->retries, result);
+ mutex_unlock(&con->sock_mutex);
+ msleep(1000);
lowcomms_connect_sock(con);
- result = 0;
+ return;
}
out:
mutex_unlock(&con->sock_mutex);
@@ -1011,16 +1265,18 @@ static struct socket *tcp_create_listen_sock(struct connection *con,
goto create_out;
}
+ /* Turn off Nagle's algorithm */
+ kernel_setsockopt(sock, SOL_TCP, TCP_NODELAY, (char *)&one,
+ sizeof(one));
+
result = kernel_setsockopt(sock, SOL_SOCKET, SO_REUSEADDR,
(char *)&one, sizeof(one));
if (result < 0) {
log_print("Failed to set SO_REUSEADDR on socket: %d", result);
}
- sock->sk->sk_user_data = con;
con->rx_action = tcp_accept_from_sock;
con->connect_action = tcp_connect_to_sock;
- con->sock = sock;
/* Bind to our port */
make_sockaddr(saddr, dlm_config.ci_tcp_port, &addr_len);
@@ -1057,7 +1313,7 @@ static void init_local(void)
int i;
dlm_local_count = 0;
- for (i = 0; i < DLM_MAX_ADDR_COUNT - 1; i++) {
+ for (i = 0; i < DLM_MAX_ADDR_COUNT; i++) {
if (dlm_our_addr(&sas, i))
break;
@@ -1102,6 +1358,7 @@ static int sctp_listen_for_all(void)
int result = -EINVAL, num = 1, i, addr_len;
struct connection *con = nodeid2con(0, GFP_NOFS);
int bufsize = NEEDED_RMEM;
+ int one = 1;
if (!con)
return -ENOMEM;
@@ -1136,6 +1393,11 @@ static int sctp_listen_for_all(void)
goto create_delsock;
}
+ result = kernel_setsockopt(sock, SOL_SCTP, SCTP_NODELAY, (char *)&one,
+ sizeof(one));
+ if (result < 0)
+ log_print("Could not set SCTP NODELAY error %d\n", result);
+
/* Init con struct */
sock->sk->sk_user_data = con;
con->sock = sock;
@@ -1230,7 +1492,6 @@ void *dlm_lowcomms_get_buffer(int nodeid, int len, gfp_t allocation, char **ppc)
struct connection *con;
struct writequeue_entry *e;
int offset = 0;
- int users = 0;
con = nodeid2con(nodeid, allocation);
if (!con)
@@ -1244,7 +1505,7 @@ void *dlm_lowcomms_get_buffer(int nodeid, int len, gfp_t allocation, char **ppc)
} else {
offset = e->end;
e->end += len;
- users = e->users++;
+ e->users++;
}
spin_unlock(&con->writequeue_lock);
@@ -1259,7 +1520,7 @@ void *dlm_lowcomms_get_buffer(int nodeid, int len, gfp_t allocation, char **ppc)
spin_lock(&con->writequeue_lock);
offset = e->end;
e->end += len;
- users = e->users++;
+ e->users++;
list_add_tail(&e->list, &con->writequeue);
spin_unlock(&con->writequeue_lock);
goto got_one;
@@ -1297,6 +1558,7 @@ static void send_to_sock(struct connection *con)
const int msg_flags = MSG_DONTWAIT | MSG_NOSIGNAL;
struct writequeue_entry *e;
int len, offset;
+ int count = 0;
mutex_lock(&con->sock_mutex);
if (con->sock == NULL)
@@ -1319,24 +1581,29 @@ static void send_to_sock(struct connection *con)
ret = kernel_sendpage(con->sock, e->page, offset, len,
msg_flags);
if (ret == -EAGAIN || ret == 0) {
+ if (ret == -EAGAIN &&
+ test_bit(SOCK_ASYNC_NOSPACE, &con->sock->flags) &&
+ !test_and_set_bit(CF_APP_LIMITED, &con->flags)) {
+ /* Notify TCP that we're limited by the
+ * application window size.
+ */
+ set_bit(SOCK_NOSPACE, &con->sock->flags);
+ con->sock->sk->sk_write_pending++;
+ }
cond_resched();
goto out;
- }
- if (ret <= 0)
+ } else if (ret < 0)
goto send_error;
}
- /* Don't starve people filling buffers */
+
+ /* Don't starve people filling buffers */
+ if (++count >= MAX_SEND_MSG_COUNT) {
cond_resched();
+ count = 0;
+ }
spin_lock(&con->writequeue_lock);
- e->offset += ret;
- e->len -= ret;
-
- if (e->len == 0 && e->users == 0) {
- list_del(&e->list);
- free_entry(e);
- continue;
- }
+ writequeue_entry_complete(e, ret);
}
spin_unlock(&con->writequeue_lock);
out:
@@ -1353,7 +1620,6 @@ out_connect:
mutex_unlock(&con->sock_mutex);
if (!test_bit(CF_INIT_PENDING, &con->flags))
lowcomms_connect_sock(con);
- return;
}
static void clean_one_writequeue(struct connection *con)
@@ -1373,6 +1639,7 @@ static void clean_one_writequeue(struct connection *con)
int dlm_lowcomms_close(int nodeid)
{
struct connection *con;
+ struct dlm_node_addr *na;
log_print("closing connection to node %d", nodeid);
con = nodeid2con(nodeid, 0);
@@ -1387,6 +1654,17 @@ int dlm_lowcomms_close(int nodeid)
clean_one_writequeue(con);
close_connection(con, true);
}
+
+ spin_lock(&dlm_node_addrs_spin);
+ na = find_node_addr(nodeid);
+ if (na) {
+ list_del(&na->list);
+ while (na->addr_count--)
+ kfree(na->addr[na->addr_count]);
+ kfree(na);
+ }
+ spin_unlock(&dlm_node_addrs_spin);
+
return 0;
}
@@ -1430,20 +1708,19 @@ static void work_stop(void)
static int work_start(void)
{
- int error;
- recv_workqueue = create_workqueue("dlm_recv");
- error = IS_ERR(recv_workqueue);
- if (error) {
- log_print("can't start dlm_recv %d", error);
- return error;
- }
-
- send_workqueue = create_singlethread_workqueue("dlm_send");
- error = IS_ERR(send_workqueue);
- if (error) {
- log_print("can't start dlm_send %d", error);
+ recv_workqueue = alloc_workqueue("dlm_recv",
+ WQ_UNBOUND | WQ_MEM_RECLAIM, 1);
+ if (!recv_workqueue) {
+ log_print("can't start dlm_recv");
+ return -ENOMEM;
+ }
+
+ send_workqueue = alloc_workqueue("dlm_send",
+ WQ_UNBOUND | WQ_MEM_RECLAIM, 1);
+ if (!send_workqueue) {
+ log_print("can't start dlm_send");
destroy_workqueue(recv_workqueue);
- return error;
+ return -ENOMEM;
}
return 0;
@@ -1471,6 +1748,7 @@ void dlm_lowcomms_stop(void)
socket activity.
*/
mutex_lock(&connections_lock);
+ dlm_allow_conn = 0;
foreach_conn(stop_conn);
mutex_unlock(&connections_lock);
@@ -1498,7 +1776,7 @@ int dlm_lowcomms_start(void)
if (!dlm_local_count) {
error = -ENOTCONN;
log_print("no local IP address has been set");
- goto out;
+ goto fail;
}
error = -ENOMEM;
@@ -1506,7 +1784,13 @@ int dlm_lowcomms_start(void)
__alignof__(struct connection), 0,
NULL);
if (!con_cache)
- goto out;
+ goto fail;
+
+ error = work_start();
+ if (error)
+ goto fail_destroy;
+
+ dlm_allow_conn = 1;
/* Start listening */
if (dlm_config.ci_protocol == 0)
@@ -1516,20 +1800,31 @@ int dlm_lowcomms_start(void)
if (error)
goto fail_unlisten;
- error = work_start();
- if (error)
- goto fail_unlisten;
-
return 0;
fail_unlisten:
+ dlm_allow_conn = 0;
con = nodeid2con(0,0);
if (con) {
close_connection(con, false);
kmem_cache_free(con_cache, con);
}
+fail_destroy:
kmem_cache_destroy(con_cache);
-
-out:
+fail:
return error;
}
+
+void dlm_lowcomms_exit(void)
+{
+ struct dlm_node_addr *na, *safe;
+
+ spin_lock(&dlm_node_addrs_spin);
+ list_for_each_entry_safe(na, safe, &dlm_node_addrs, list) {
+ list_del(&na->list);
+ while (na->addr_count--)
+ kfree(na->addr[na->addr_count]);
+ kfree(na);
+ }
+ spin_unlock(&dlm_node_addrs_spin);
+}
diff --git a/fs/dlm/lowcomms.h b/fs/dlm/lowcomms.h
index 1311e642628..67462e54fc2 100644
--- a/fs/dlm/lowcomms.h
+++ b/fs/dlm/lowcomms.h
@@ -16,10 +16,12 @@
int dlm_lowcomms_start(void);
void dlm_lowcomms_stop(void);
+void dlm_lowcomms_exit(void);
int dlm_lowcomms_close(int nodeid);
void *dlm_lowcomms_get_buffer(int nodeid, int len, gfp_t allocation, char **ppc);
void dlm_lowcomms_commit_buffer(void *mh);
int dlm_lowcomms_connect_node(int nodeid);
+int dlm_lowcomms_addr(int nodeid, struct sockaddr_storage *addr, int len);
#endif /* __LOWCOMMS_DOT_H__ */
diff --git a/fs/dlm/main.c b/fs/dlm/main.c
index b80e0aa3cfa..079c0bd71ab 100644
--- a/fs/dlm/main.c
+++ b/fs/dlm/main.c
@@ -17,6 +17,7 @@
#include "user.h"
#include "memory.h"
#include "config.h"
+#include "lowcomms.h"
static int __init init_dlm(void)
{
@@ -50,7 +51,7 @@ static int __init init_dlm(void)
if (error)
goto out_netlink;
- printk("DLM (built %s %s) installed\n", __DATE__, __TIME__);
+ printk("DLM installed\n");
return 0;
@@ -78,6 +79,7 @@ static void __exit exit_dlm(void)
dlm_config_exit();
dlm_memory_exit();
dlm_lockspace_exit();
+ dlm_lowcomms_exit();
dlm_unregister_debugfs();
}
diff --git a/fs/dlm/member.c b/fs/dlm/member.c
index b12532e553f..9c47f1c14a8 100644
--- a/fs/dlm/member.c
+++ b/fs/dlm/member.c
@@ -1,7 +1,7 @@
/******************************************************************************
*******************************************************************************
**
-** Copyright (C) 2005-2009 Red Hat, Inc. All rights reserved.
+** Copyright (C) 2005-2011 Red Hat, Inc. All rights reserved.
**
** This copyrighted material is made available to anyone wishing to use,
** modify, copy, or redistribute it subject to the terms and conditions
@@ -19,6 +19,277 @@
#include "config.h"
#include "lowcomms.h"
+int dlm_slots_version(struct dlm_header *h)
+{
+ if ((h->h_version & 0x0000FFFF) < DLM_HEADER_SLOTS)
+ return 0;
+ return 1;
+}
+
+void dlm_slot_save(struct dlm_ls *ls, struct dlm_rcom *rc,
+ struct dlm_member *memb)
+{
+ struct rcom_config *rf = (struct rcom_config *)rc->rc_buf;
+
+ if (!dlm_slots_version(&rc->rc_header))
+ return;
+
+ memb->slot = le16_to_cpu(rf->rf_our_slot);
+ memb->generation = le32_to_cpu(rf->rf_generation);
+}
+
+void dlm_slots_copy_out(struct dlm_ls *ls, struct dlm_rcom *rc)
+{
+ struct dlm_slot *slot;
+ struct rcom_slot *ro;
+ int i;
+
+ ro = (struct rcom_slot *)(rc->rc_buf + sizeof(struct rcom_config));
+
+ /* ls_slots array is sparse, but not rcom_slots */
+
+ for (i = 0; i < ls->ls_slots_size; i++) {
+ slot = &ls->ls_slots[i];
+ if (!slot->nodeid)
+ continue;
+ ro->ro_nodeid = cpu_to_le32(slot->nodeid);
+ ro->ro_slot = cpu_to_le16(slot->slot);
+ ro++;
+ }
+}
+
+#define SLOT_DEBUG_LINE 128
+
+static void log_slots(struct dlm_ls *ls, uint32_t gen, int num_slots,
+ struct rcom_slot *ro0, struct dlm_slot *array,
+ int array_size)
+{
+ char line[SLOT_DEBUG_LINE];
+ int len = SLOT_DEBUG_LINE - 1;
+ int pos = 0;
+ int ret, i;
+
+ memset(line, 0, sizeof(line));
+
+ if (array) {
+ for (i = 0; i < array_size; i++) {
+ if (!array[i].nodeid)
+ continue;
+
+ ret = snprintf(line + pos, len - pos, " %d:%d",
+ array[i].slot, array[i].nodeid);
+ if (ret >= len - pos)
+ break;
+ pos += ret;
+ }
+ } else if (ro0) {
+ for (i = 0; i < num_slots; i++) {
+ ret = snprintf(line + pos, len - pos, " %d:%d",
+ ro0[i].ro_slot, ro0[i].ro_nodeid);
+ if (ret >= len - pos)
+ break;
+ pos += ret;
+ }
+ }
+
+ log_rinfo(ls, "generation %u slots %d%s", gen, num_slots, line);
+}
+
+int dlm_slots_copy_in(struct dlm_ls *ls)
+{
+ struct dlm_member *memb;
+ struct dlm_rcom *rc = ls->ls_recover_buf;
+ struct rcom_config *rf = (struct rcom_config *)rc->rc_buf;
+ struct rcom_slot *ro0, *ro;
+ int our_nodeid = dlm_our_nodeid();
+ int i, num_slots;
+ uint32_t gen;
+
+ if (!dlm_slots_version(&rc->rc_header))
+ return -1;
+
+ gen = le32_to_cpu(rf->rf_generation);
+ if (gen <= ls->ls_generation) {
+ log_error(ls, "dlm_slots_copy_in gen %u old %u",
+ gen, ls->ls_generation);
+ }
+ ls->ls_generation = gen;
+
+ num_slots = le16_to_cpu(rf->rf_num_slots);
+ if (!num_slots)
+ return -1;
+
+ ro0 = (struct rcom_slot *)(rc->rc_buf + sizeof(struct rcom_config));
+
+ for (i = 0, ro = ro0; i < num_slots; i++, ro++) {
+ ro->ro_nodeid = le32_to_cpu(ro->ro_nodeid);
+ ro->ro_slot = le16_to_cpu(ro->ro_slot);
+ }
+
+ log_slots(ls, gen, num_slots, ro0, NULL, 0);
+
+ list_for_each_entry(memb, &ls->ls_nodes, list) {
+ for (i = 0, ro = ro0; i < num_slots; i++, ro++) {
+ if (ro->ro_nodeid != memb->nodeid)
+ continue;
+ memb->slot = ro->ro_slot;
+ memb->slot_prev = memb->slot;
+ break;
+ }
+
+ if (memb->nodeid == our_nodeid) {
+ if (ls->ls_slot && ls->ls_slot != memb->slot) {
+ log_error(ls, "dlm_slots_copy_in our slot "
+ "changed %d %d", ls->ls_slot,
+ memb->slot);
+ return -1;
+ }
+
+ if (!ls->ls_slot)
+ ls->ls_slot = memb->slot;
+ }
+
+ if (!memb->slot) {
+ log_error(ls, "dlm_slots_copy_in nodeid %d no slot",
+ memb->nodeid);
+ return -1;
+ }
+ }
+
+ return 0;
+}
+
+/* for any nodes that do not support slots, we will not have set memb->slot
+ in wait_status_all(), so memb->slot will remain -1, and we will not
+ assign slots or set ls_num_slots here */
+
+int dlm_slots_assign(struct dlm_ls *ls, int *num_slots, int *slots_size,
+ struct dlm_slot **slots_out, uint32_t *gen_out)
+{
+ struct dlm_member *memb;
+ struct dlm_slot *array;
+ int our_nodeid = dlm_our_nodeid();
+ int array_size, max_slots, i;
+ int need = 0;
+ int max = 0;
+ int num = 0;
+ uint32_t gen = 0;
+
+ /* our own memb struct will have slot -1 gen 0 */
+
+ list_for_each_entry(memb, &ls->ls_nodes, list) {
+ if (memb->nodeid == our_nodeid) {
+ memb->slot = ls->ls_slot;
+ memb->generation = ls->ls_generation;
+ break;
+ }
+ }
+
+ list_for_each_entry(memb, &ls->ls_nodes, list) {
+ if (memb->generation > gen)
+ gen = memb->generation;
+
+ /* node doesn't support slots */
+
+ if (memb->slot == -1)
+ return -1;
+
+ /* node needs a slot assigned */
+
+ if (!memb->slot)
+ need++;
+
+ /* node has a slot assigned */
+
+ num++;
+
+ if (!max || max < memb->slot)
+ max = memb->slot;
+
+ /* sanity check, once slot is assigned it shouldn't change */
+
+ if (memb->slot_prev && memb->slot && memb->slot_prev != memb->slot) {
+ log_error(ls, "nodeid %d slot changed %d %d",
+ memb->nodeid, memb->slot_prev, memb->slot);
+ return -1;
+ }
+ memb->slot_prev = memb->slot;
+ }
+
+ array_size = max + need;
+
+ array = kzalloc(array_size * sizeof(struct dlm_slot), GFP_NOFS);
+ if (!array)
+ return -ENOMEM;
+
+ num = 0;
+
+ /* fill in slots (offsets) that are used */
+
+ list_for_each_entry(memb, &ls->ls_nodes, list) {
+ if (!memb->slot)
+ continue;
+
+ if (memb->slot > array_size) {
+ log_error(ls, "invalid slot number %d", memb->slot);
+ kfree(array);
+ return -1;
+ }
+
+ array[memb->slot - 1].nodeid = memb->nodeid;
+ array[memb->slot - 1].slot = memb->slot;
+ num++;
+ }
+
+ /* assign new slots from unused offsets */
+
+ list_for_each_entry(memb, &ls->ls_nodes, list) {
+ if (memb->slot)
+ continue;
+
+ for (i = 0; i < array_size; i++) {
+ if (array[i].nodeid)
+ continue;
+
+ memb->slot = i + 1;
+ memb->slot_prev = memb->slot;
+ array[i].nodeid = memb->nodeid;
+ array[i].slot = memb->slot;
+ num++;
+
+ if (!ls->ls_slot && memb->nodeid == our_nodeid)
+ ls->ls_slot = memb->slot;
+ break;
+ }
+
+ if (!memb->slot) {
+ log_error(ls, "no free slot found");
+ kfree(array);
+ return -1;
+ }
+ }
+
+ gen++;
+
+ log_slots(ls, gen, num, NULL, array, array_size);
+
+ max_slots = (dlm_config.ci_buffer_size - sizeof(struct dlm_rcom) -
+ sizeof(struct rcom_config)) / sizeof(struct rcom_slot);
+
+ if (num > max_slots) {
+ log_error(ls, "num_slots %d exceeds max_slots %d",
+ num, max_slots);
+ kfree(array);
+ return -1;
+ }
+
+ *gen_out = gen;
+ *slots_out = array;
+ *slots_size = array_size;
+ *num_slots = num;
+ return 0;
+}
+
static void add_ordered_member(struct dlm_ls *ls, struct dlm_member *new)
{
struct dlm_member *memb = NULL;
@@ -43,59 +314,51 @@ static void add_ordered_member(struct dlm_ls *ls, struct dlm_member *new)
}
}
-static int dlm_add_member(struct dlm_ls *ls, int nodeid)
+static int dlm_add_member(struct dlm_ls *ls, struct dlm_config_node *node)
{
struct dlm_member *memb;
- int w, error;
+ int error;
memb = kzalloc(sizeof(struct dlm_member), GFP_NOFS);
if (!memb)
return -ENOMEM;
- w = dlm_node_weight(ls->ls_name, nodeid);
- if (w < 0) {
- kfree(memb);
- return w;
- }
-
- error = dlm_lowcomms_connect_node(nodeid);
+ error = dlm_lowcomms_connect_node(node->nodeid);
if (error < 0) {
kfree(memb);
return error;
}
- memb->nodeid = nodeid;
- memb->weight = w;
+ memb->nodeid = node->nodeid;
+ memb->weight = node->weight;
+ memb->comm_seq = node->comm_seq;
add_ordered_member(ls, memb);
ls->ls_num_nodes++;
return 0;
}
-static void dlm_remove_member(struct dlm_ls *ls, struct dlm_member *memb)
-{
- list_move(&memb->list, &ls->ls_nodes_gone);
- ls->ls_num_nodes--;
-}
-
-int dlm_is_member(struct dlm_ls *ls, int nodeid)
+static struct dlm_member *find_memb(struct list_head *head, int nodeid)
{
struct dlm_member *memb;
- list_for_each_entry(memb, &ls->ls_nodes, list) {
+ list_for_each_entry(memb, head, list) {
if (memb->nodeid == nodeid)
- return 1;
+ return memb;
}
+ return NULL;
+}
+
+int dlm_is_member(struct dlm_ls *ls, int nodeid)
+{
+ if (find_memb(&ls->ls_nodes, nodeid))
+ return 1;
return 0;
}
int dlm_is_removed(struct dlm_ls *ls, int nodeid)
{
- struct dlm_member *memb;
-
- list_for_each_entry(memb, &ls->ls_nodes_gone, list) {
- if (memb->nodeid == nodeid)
- return 1;
- }
+ if (find_memb(&ls->ls_nodes_gone, nodeid))
+ return 1;
return 0;
}
@@ -176,72 +439,136 @@ static int ping_members(struct dlm_ls *ls)
error = dlm_recovery_stopped(ls);
if (error)
break;
- error = dlm_rcom_status(ls, memb->nodeid);
+ error = dlm_rcom_status(ls, memb->nodeid, 0);
if (error)
break;
}
if (error)
- log_debug(ls, "ping_members aborted %d last nodeid %d",
+ log_rinfo(ls, "ping_members aborted %d last nodeid %d",
error, ls->ls_recover_nodeid);
return error;
}
+static void dlm_lsop_recover_prep(struct dlm_ls *ls)
+{
+ if (!ls->ls_ops || !ls->ls_ops->recover_prep)
+ return;
+ ls->ls_ops->recover_prep(ls->ls_ops_arg);
+}
+
+static void dlm_lsop_recover_slot(struct dlm_ls *ls, struct dlm_member *memb)
+{
+ struct dlm_slot slot;
+ uint32_t seq;
+ int error;
+
+ if (!ls->ls_ops || !ls->ls_ops->recover_slot)
+ return;
+
+ /* if there is no comms connection with this node
+ or the present comms connection is newer
+ than the one when this member was added, then
+ we consider the node to have failed (versus
+ being removed due to dlm_release_lockspace) */
+
+ error = dlm_comm_seq(memb->nodeid, &seq);
+
+ if (!error && seq == memb->comm_seq)
+ return;
+
+ slot.nodeid = memb->nodeid;
+ slot.slot = memb->slot;
+
+ ls->ls_ops->recover_slot(ls->ls_ops_arg, &slot);
+}
+
+void dlm_lsop_recover_done(struct dlm_ls *ls)
+{
+ struct dlm_member *memb;
+ struct dlm_slot *slots;
+ int i, num;
+
+ if (!ls->ls_ops || !ls->ls_ops->recover_done)
+ return;
+
+ num = ls->ls_num_nodes;
+
+ slots = kzalloc(num * sizeof(struct dlm_slot), GFP_KERNEL);
+ if (!slots)
+ return;
+
+ i = 0;
+ list_for_each_entry(memb, &ls->ls_nodes, list) {
+ if (i == num) {
+ log_error(ls, "dlm_lsop_recover_done bad num %d", num);
+ goto out;
+ }
+ slots[i].nodeid = memb->nodeid;
+ slots[i].slot = memb->slot;
+ i++;
+ }
+
+ ls->ls_ops->recover_done(ls->ls_ops_arg, slots, num,
+ ls->ls_slot, ls->ls_generation);
+ out:
+ kfree(slots);
+}
+
+static struct dlm_config_node *find_config_node(struct dlm_recover *rv,
+ int nodeid)
+{
+ int i;
+
+ for (i = 0; i < rv->nodes_count; i++) {
+ if (rv->nodes[i].nodeid == nodeid)
+ return &rv->nodes[i];
+ }
+ return NULL;
+}
+
int dlm_recover_members(struct dlm_ls *ls, struct dlm_recover *rv, int *neg_out)
{
struct dlm_member *memb, *safe;
- int i, error, found, pos = 0, neg = 0, low = -1;
+ struct dlm_config_node *node;
+ int i, error, neg = 0, low = -1;
/* previously removed members that we've not finished removing need to
count as a negative change so the "neg" recovery steps will happen */
list_for_each_entry(memb, &ls->ls_nodes_gone, list) {
- log_debug(ls, "prev removed member %d", memb->nodeid);
+ log_rinfo(ls, "prev removed member %d", memb->nodeid);
neg++;
}
/* move departed members from ls_nodes to ls_nodes_gone */
list_for_each_entry_safe(memb, safe, &ls->ls_nodes, list) {
- found = 0;
- for (i = 0; i < rv->node_count; i++) {
- if (memb->nodeid == rv->nodeids[i]) {
- found = 1;
- break;
- }
- }
+ node = find_config_node(rv, memb->nodeid);
+ if (node && !node->new)
+ continue;
- if (!found) {
- neg++;
- dlm_remove_member(ls, memb);
- log_debug(ls, "remove member %d", memb->nodeid);
+ if (!node) {
+ log_rinfo(ls, "remove member %d", memb->nodeid);
+ } else {
+ /* removed and re-added */
+ log_rinfo(ls, "remove member %d comm_seq %u %u",
+ memb->nodeid, memb->comm_seq, node->comm_seq);
}
- }
- /* Add an entry to ls_nodes_gone for members that were removed and
- then added again, so that previous state for these nodes will be
- cleared during recovery. */
-
- for (i = 0; i < rv->new_count; i++) {
- if (!dlm_is_member(ls, rv->new[i]))
- continue;
- log_debug(ls, "new nodeid %d is a re-added member", rv->new[i]);
-
- memb = kzalloc(sizeof(struct dlm_member), GFP_NOFS);
- if (!memb)
- return -ENOMEM;
- memb->nodeid = rv->new[i];
- list_add_tail(&memb->list, &ls->ls_nodes_gone);
neg++;
+ list_move(&memb->list, &ls->ls_nodes_gone);
+ ls->ls_num_nodes--;
+ dlm_lsop_recover_slot(ls, memb);
}
/* add new members to ls_nodes */
- for (i = 0; i < rv->node_count; i++) {
- if (dlm_is_member(ls, rv->nodeids[i]))
+ for (i = 0; i < rv->nodes_count; i++) {
+ node = &rv->nodes[i];
+ if (dlm_is_member(ls, node->nodeid))
continue;
- dlm_add_member(ls, rv->nodeids[i]);
- pos++;
- log_debug(ls, "add member %d", rv->nodeids[i]);
+ dlm_add_member(ls, node);
+ log_rinfo(ls, "add member %d", node->nodeid);
}
list_for_each_entry(memb, &ls->ls_nodes, list) {
@@ -251,7 +578,6 @@ int dlm_recover_members(struct dlm_ls *ls, struct dlm_recover *rv, int *neg_out)
ls->ls_low_nodeid = low;
make_member_array(ls);
- dlm_set_recover_status(ls, DLM_RS_NODES);
*neg_out = neg;
error = ping_members(ls);
@@ -261,12 +587,8 @@ int dlm_recover_members(struct dlm_ls *ls, struct dlm_recover *rv, int *neg_out)
ls->ls_members_result = error;
complete(&ls->ls_members_done);
}
- if (error)
- goto out;
- error = dlm_recover_members_wait(ls);
- out:
- log_debug(ls, "total members %d error %d", ls->ls_num_nodes, error);
+ log_rinfo(ls, "dlm_recover_members %d nodes", ls->ls_num_nodes);
return error;
}
@@ -291,13 +613,13 @@ int dlm_ls_stop(struct dlm_ls *ls)
down_write(&ls->ls_recv_active);
/*
- * Abort any recovery that's in progress (see RECOVERY_STOP,
+ * Abort any recovery that's in progress (see RECOVER_STOP,
* dlm_recovery_stopped()) and tell any other threads running in the
* dlm to quit any processing (see RUNNING, dlm_locking_stopped()).
*/
spin_lock(&ls->ls_recover_lock);
- set_bit(LSFL_RECOVERY_STOP, &ls->ls_flags);
+ set_bit(LSFL_RECOVER_STOP, &ls->ls_flags);
new = test_and_clear_bit(LSFL_RUNNING, &ls->ls_flags);
ls->ls_recover_seq++;
spin_unlock(&ls->ls_recover_lock);
@@ -317,36 +639,49 @@ int dlm_ls_stop(struct dlm_ls *ls)
* when recovery is complete.
*/
- if (new)
- down_write(&ls->ls_in_recovery);
+ if (new) {
+ set_bit(LSFL_RECOVER_DOWN, &ls->ls_flags);
+ wake_up_process(ls->ls_recoverd_task);
+ wait_event(ls->ls_recover_lock_wait,
+ test_bit(LSFL_RECOVER_LOCK, &ls->ls_flags));
+ }
/*
* The recoverd suspend/resume makes sure that dlm_recoverd (if
- * running) has noticed RECOVERY_STOP above and quit processing the
+ * running) has noticed RECOVER_STOP above and quit processing the
* previous recovery.
*/
dlm_recoverd_suspend(ls);
+
+ spin_lock(&ls->ls_recover_lock);
+ kfree(ls->ls_slots);
+ ls->ls_slots = NULL;
+ ls->ls_num_slots = 0;
+ ls->ls_slots_size = 0;
ls->ls_recover_status = 0;
+ spin_unlock(&ls->ls_recover_lock);
+
dlm_recoverd_resume(ls);
if (!ls->ls_recover_begin)
ls->ls_recover_begin = jiffies;
+
+ dlm_lsop_recover_prep(ls);
return 0;
}
int dlm_ls_start(struct dlm_ls *ls)
{
struct dlm_recover *rv = NULL, *rv_old;
- int *ids = NULL, *new = NULL;
- int error, ids_count = 0, new_count = 0;
+ struct dlm_config_node *nodes;
+ int error, count;
rv = kzalloc(sizeof(struct dlm_recover), GFP_NOFS);
if (!rv)
return -ENOMEM;
- error = dlm_nodeid_list(ls->ls_name, &ids, &ids_count,
- &new, &new_count);
+ error = dlm_config_nodes(ls->ls_name, &nodes, &count);
if (error < 0)
goto fail;
@@ -361,10 +696,8 @@ int dlm_ls_start(struct dlm_ls *ls)
goto fail;
}
- rv->nodeids = ids;
- rv->node_count = ids_count;
- rv->new = new;
- rv->new_count = new_count;
+ rv->nodes = nodes;
+ rv->nodes_count = count;
rv->seq = ++ls->ls_recover_seq;
rv_old = ls->ls_recover_args;
ls->ls_recover_args = rv;
@@ -372,19 +705,18 @@ int dlm_ls_start(struct dlm_ls *ls)
if (rv_old) {
log_error(ls, "unused recovery %llx %d",
- (unsigned long long)rv_old->seq, rv_old->node_count);
- kfree(rv_old->nodeids);
- kfree(rv_old->new);
+ (unsigned long long)rv_old->seq, rv_old->nodes_count);
+ kfree(rv_old->nodes);
kfree(rv_old);
}
- dlm_recoverd_kick(ls);
+ set_bit(LSFL_RECOVER_WORK, &ls->ls_flags);
+ wake_up_process(ls->ls_recoverd_task);
return 0;
fail:
kfree(rv);
- kfree(ids);
- kfree(new);
+ kfree(nodes);
return error;
}
diff --git a/fs/dlm/member.h b/fs/dlm/member.h
index 7a26fca1e0b..3deb70661c6 100644
--- a/fs/dlm/member.h
+++ b/fs/dlm/member.h
@@ -1,7 +1,7 @@
/******************************************************************************
*******************************************************************************
**
-** Copyright (C) 2005-2008 Red Hat, Inc. All rights reserved.
+** Copyright (C) 2005-2011 Red Hat, Inc. All rights reserved.
**
** This copyrighted material is made available to anyone wishing to use,
** modify, copy, or redistribute it subject to the terms and conditions
@@ -20,6 +20,14 @@ 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);
+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);
+int dlm_slots_copy_in(struct dlm_ls *ls);
+int dlm_slots_assign(struct dlm_ls *ls, int *num_slots, int *slots_size,
+ struct dlm_slot **slots_out, uint32_t *gen_out);
+void dlm_lsop_recover_done(struct dlm_ls *ls);
#endif /* __MEMBER_DOT_H__ */
diff --git a/fs/dlm/memory.c b/fs/dlm/memory.c
index 8e0d00db004..7cd24bccd4f 100644
--- a/fs/dlm/memory.c
+++ b/fs/dlm/memory.c
@@ -16,23 +16,32 @@
#include "memory.h"
static struct kmem_cache *lkb_cache;
+static struct kmem_cache *rsb_cache;
int __init dlm_memory_init(void)
{
- int ret = 0;
-
lkb_cache = kmem_cache_create("dlm_lkb", sizeof(struct dlm_lkb),
__alignof__(struct dlm_lkb), 0, NULL);
if (!lkb_cache)
- ret = -ENOMEM;
- return ret;
+ return -ENOMEM;
+
+ rsb_cache = kmem_cache_create("dlm_rsb", sizeof(struct dlm_rsb),
+ __alignof__(struct dlm_rsb), 0, NULL);
+ if (!rsb_cache) {
+ kmem_cache_destroy(lkb_cache);
+ return -ENOMEM;
+ }
+
+ return 0;
}
void dlm_memory_exit(void)
{
if (lkb_cache)
kmem_cache_destroy(lkb_cache);
+ if (rsb_cache)
+ kmem_cache_destroy(rsb_cache);
}
char *dlm_allocate_lvb(struct dlm_ls *ls)
@@ -48,16 +57,11 @@ void dlm_free_lvb(char *p)
kfree(p);
}
-/* FIXME: have some minimal space built-in to rsb for the name and
- kmalloc a separate name if needed, like dentries are done */
-
-struct dlm_rsb *dlm_allocate_rsb(struct dlm_ls *ls, int namelen)
+struct dlm_rsb *dlm_allocate_rsb(struct dlm_ls *ls)
{
struct dlm_rsb *r;
- DLM_ASSERT(namelen <= DLM_RESNAME_MAXLEN,);
-
- r = kzalloc(sizeof(*r) + namelen, GFP_NOFS);
+ r = kmem_cache_zalloc(rsb_cache, GFP_NOFS);
return r;
}
@@ -65,7 +69,7 @@ void dlm_free_rsb(struct dlm_rsb *r)
{
if (r->res_lvbptr)
dlm_free_lvb(r->res_lvbptr);
- kfree(r);
+ kmem_cache_free(rsb_cache, r);
}
struct dlm_lkb *dlm_allocate_lkb(struct dlm_ls *ls)
diff --git a/fs/dlm/memory.h b/fs/dlm/memory.h
index 485fb29143b..177c11cbb0a 100644
--- a/fs/dlm/memory.h
+++ b/fs/dlm/memory.h
@@ -16,7 +16,7 @@
int dlm_memory_init(void);
void dlm_memory_exit(void);
-struct dlm_rsb *dlm_allocate_rsb(struct dlm_ls *ls, int namelen);
+struct dlm_rsb *dlm_allocate_rsb(struct dlm_ls *ls);
void dlm_free_rsb(struct dlm_rsb *r);
struct dlm_lkb *dlm_allocate_lkb(struct dlm_ls *ls);
void dlm_free_lkb(struct dlm_lkb *l);
diff --git a/fs/dlm/netlink.c b/fs/dlm/netlink.c
index ef17e0169da..e7cfbaf8d0e 100644
--- a/fs/dlm/netlink.c
+++ b/fs/dlm/netlink.c
@@ -14,7 +14,7 @@
#include "dlm_internal.h"
static uint32_t dlm_nl_seqnum;
-static uint32_t listener_nlpid;
+static uint32_t listener_nlportid;
static struct genl_family family = {
.id = GENL_ID_GENERATE,
@@ -64,24 +64,26 @@ static int send_data(struct sk_buff *skb)
return rv;
}
- return genlmsg_unicast(&init_net, skb, listener_nlpid);
+ return genlmsg_unicast(&init_net, skb, listener_nlportid);
}
static int user_cmd(struct sk_buff *skb, struct genl_info *info)
{
- listener_nlpid = info->snd_pid;
- printk("user_cmd nlpid %u\n", listener_nlpid);
+ listener_nlportid = info->snd_portid;
+ printk("user_cmd nlpid %u\n", listener_nlportid);
return 0;
}
-static struct genl_ops dlm_nl_ops = {
- .cmd = DLM_CMD_HELLO,
- .doit = user_cmd,
+static struct genl_ops dlm_nl_ops[] = {
+ {
+ .cmd = DLM_CMD_HELLO,
+ .doit = user_cmd,
+ },
};
int __init dlm_netlink_init(void)
{
- return genl_register_family_with_ops(&family, &dlm_nl_ops, 1);
+ return genl_register_family_with_ops(&family, dlm_nl_ops);
}
void dlm_netlink_exit(void)
diff --git a/fs/dlm/plock.c b/fs/dlm/plock.c
index 30d8b85febb..f704458ea5f 100644
--- a/fs/dlm/plock.c
+++ b/fs/dlm/plock.c
@@ -71,6 +71,36 @@ 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(struct dlm_ls *ls, u64 number,
+ struct file *file, struct file_lock *fl)
+{
+ struct plock_op *op;
+
+ op = kzalloc(sizeof(*op), GFP_NOFS);
+ if (!op)
+ return;
+
+ op->info.optype = DLM_PLOCK_OP_UNLOCK;
+ op->info.pid = fl->fl_pid;
+ op->info.fsid = ls->ls_global_id;
+ op->info.number = number;
+ op->info.start = 0;
+ op->info.end = OFFSET_MAX;
+ if (fl->fl_lmops && fl->fl_lmops->lm_grant)
+ op->info.owner = (__u64) fl->fl_pid;
+ else
+ op->info.owner = (__u64)(long) fl->fl_owner;
+
+ op->info.flags |= DLM_PLOCK_FL_CLOSE;
+ send_op(op);
+}
+
int dlm_posix_lock(dlm_lockspace_t *lockspace, u64 number, struct file *file,
int cmd, struct file_lock *fl)
{
@@ -98,11 +128,11 @@ int dlm_posix_lock(dlm_lockspace_t *lockspace, u64 number, struct file *file,
op->info.number = number;
op->info.start = fl->fl_start;
op->info.end = fl->fl_end;
- if (fl->fl_lmops && fl->fl_lmops->fl_grant) {
+ if (fl->fl_lmops && fl->fl_lmops->lm_grant) {
/* fl_owner is lockd which doesn't distinguish
processes on the nfs client */
op->info.owner = (__u64) fl->fl_pid;
- xop->callback = fl->fl_lmops->fl_grant;
+ xop->callback = fl->fl_lmops->lm_grant;
locks_init_lock(&xop->flc);
locks_copy_lock(&xop->flc, fl);
xop->fl = fl;
@@ -114,9 +144,19 @@ int dlm_posix_lock(dlm_lockspace_t *lockspace, u64 number, struct file *file,
send_op(op);
- if (xop->callback == NULL)
- wait_event(recv_wq, (op->done != 0));
- else {
+ if (xop->callback == NULL) {
+ rv = wait_event_killable(recv_wq, (op->done != 0));
+ if (rv == -ERESTARTSYS) {
+ log_debug(ls, "dlm_posix_lock: wait killed %llx",
+ (unsigned long long)number);
+ spin_lock(&ops_lock);
+ list_del(&op->list);
+ spin_unlock(&ops_lock);
+ kfree(xop);
+ do_unlock_close(ls, number, file, fl);
+ goto out;
+ }
+ } else {
rv = FILE_LOCK_DEFERRED;
goto out;
}
@@ -207,6 +247,7 @@ int dlm_posix_unlock(dlm_lockspace_t *lockspace, u64 number, struct file *file,
struct dlm_ls *ls;
struct plock_op *op;
int rv;
+ unsigned char fl_flags = fl->fl_flags;
ls = dlm_find_lockspace_local(lockspace);
if (!ls)
@@ -218,9 +259,18 @@ int dlm_posix_unlock(dlm_lockspace_t *lockspace, u64 number, struct file *file,
goto out;
}
- if (posix_lock_file_wait(file, fl) < 0)
- log_error(ls, "dlm_posix_unlock: vfs unlock error %llx",
- (unsigned long long)number);
+ /* cause the vfs unlock to return ENOENT if lock is not found */
+ fl->fl_flags |= FL_EXISTS;
+
+ rv = posix_lock_file_wait(file, fl);
+ if (rv == -ENOENT) {
+ rv = 0;
+ goto out_free;
+ }
+ if (rv < 0) {
+ log_error(ls, "dlm_posix_unlock: vfs unlock error %d %llx",
+ rv, (unsigned long long)number);
+ }
op->info.optype = DLM_PLOCK_OP_UNLOCK;
op->info.pid = fl->fl_pid;
@@ -228,11 +278,18 @@ int dlm_posix_unlock(dlm_lockspace_t *lockspace, u64 number, struct file *file,
op->info.number = number;
op->info.start = fl->fl_start;
op->info.end = fl->fl_end;
- if (fl->fl_lmops && fl->fl_lmops->fl_grant)
+ if (fl->fl_lmops && fl->fl_lmops->lm_grant)
op->info.owner = (__u64) fl->fl_pid;
else
op->info.owner = (__u64)(long) fl->fl_owner;
+ if (fl->fl_flags & FL_CLOSE) {
+ op->info.flags |= DLM_PLOCK_FL_CLOSE;
+ send_op(op);
+ rv = 0;
+ goto out;
+ }
+
send_op(op);
wait_event(recv_wq, (op->done != 0));
@@ -249,9 +306,11 @@ int dlm_posix_unlock(dlm_lockspace_t *lockspace, u64 number, struct file *file,
if (rv == -ENOENT)
rv = 0;
+out_free:
kfree(op);
out:
dlm_put_lockspace(ls);
+ fl->fl_flags = fl_flags;
return rv;
}
EXPORT_SYMBOL_GPL(dlm_posix_unlock);
@@ -280,7 +339,7 @@ int dlm_posix_get(dlm_lockspace_t *lockspace, u64 number, struct file *file,
op->info.number = number;
op->info.start = fl->fl_start;
op->info.end = fl->fl_end;
- if (fl->fl_lmops && fl->fl_lmops->fl_grant)
+ if (fl->fl_lmops && fl->fl_lmops->lm_grant)
op->info.owner = (__u64) fl->fl_pid;
else
op->info.owner = (__u64)(long) fl->fl_owner;
@@ -334,7 +393,10 @@ static ssize_t dev_read(struct file *file, char __user *u, size_t count,
spin_lock(&ops_lock);
if (!list_empty(&send_list)) {
op = list_entry(send_list.next, struct plock_op, list);
- list_move(&op->list, &recv_list);
+ if (op->info.flags & DLM_PLOCK_FL_CLOSE)
+ list_del(&op->list);
+ else
+ list_move(&op->list, &recv_list);
memcpy(&info, &op->info, sizeof(info));
}
spin_unlock(&ops_lock);
@@ -342,6 +404,13 @@ static ssize_t dev_read(struct file *file, char __user *u, size_t count,
if (!op)
return -EAGAIN;
+ /* 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). */
+
+ if (op->info.flags & DLM_PLOCK_FL_CLOSE)
+ kfree(op);
+
if (copy_to_user(u, &info, sizeof(info)))
return -EFAULT;
return sizeof(info);
diff --git a/fs/dlm/rcom.c b/fs/dlm/rcom.c
index 3c83a49a48a..9d61947d473 100644
--- a/fs/dlm/rcom.c
+++ b/fs/dlm/rcom.c
@@ -24,7 +24,6 @@
#include "lock.h"
#include "util.h"
-
static int rcom_response(struct dlm_ls *ls)
{
return test_bit(LSFL_RCOM_READY, &ls->ls_flags);
@@ -72,20 +71,30 @@ static void send_rcom(struct dlm_ls *ls, struct dlm_mhandle *mh,
dlm_lowcomms_commit_buffer(mh);
}
+static void set_rcom_status(struct dlm_ls *ls, struct rcom_status *rs,
+ uint32_t flags)
+{
+ rs->rs_flags = cpu_to_le32(flags);
+}
+
/* When replying to a status request, a node also sends back its
configuration values. The requesting node then checks that the remote
node is configured the same way as itself. */
-static void make_config(struct dlm_ls *ls, struct rcom_config *rf)
+static void set_rcom_config(struct dlm_ls *ls, struct rcom_config *rf,
+ uint32_t num_slots)
{
rf->rf_lvblen = cpu_to_le32(ls->ls_lvblen);
rf->rf_lsflags = cpu_to_le32(ls->ls_exflags);
+
+ rf->rf_our_slot = cpu_to_le16(ls->ls_slot);
+ rf->rf_num_slots = cpu_to_le16(num_slots);
+ rf->rf_generation = cpu_to_le32(ls->ls_generation);
}
-static int check_config(struct dlm_ls *ls, struct dlm_rcom *rc, int nodeid)
+static int check_rcom_config(struct dlm_ls *ls, struct dlm_rcom *rc, int nodeid)
{
struct rcom_config *rf = (struct rcom_config *) rc->rc_buf;
- size_t conf_size = sizeof(struct dlm_rcom) + sizeof(struct rcom_config);
if ((rc->rc_header.h_version & 0xFFFF0000) != DLM_HEADER_MAJOR) {
log_error(ls, "version mismatch: %x nodeid %d: %x",
@@ -94,12 +103,6 @@ static int check_config(struct dlm_ls *ls, struct dlm_rcom *rc, int nodeid)
return -EPROTO;
}
- if (rc->rc_header.h_length < conf_size) {
- log_error(ls, "config too short: %d nodeid %d",
- rc->rc_header.h_length, nodeid);
- return -EPROTO;
- }
-
if (le32_to_cpu(rf->rf_lvblen) != ls->ls_lvblen ||
le32_to_cpu(rf->rf_lsflags) != ls->ls_exflags) {
log_error(ls, "config mismatch: %d,%x nodeid %d: %d,%x",
@@ -127,7 +130,18 @@ static void disallow_sync_reply(struct dlm_ls *ls)
spin_unlock(&ls->ls_rcom_spin);
}
-int dlm_rcom_status(struct dlm_ls *ls, int nodeid)
+/*
+ * low nodeid gathers one slot value at a time from each node.
+ * it sets need_slots=0, and saves rf_our_slot returned from each
+ * rcom_config.
+ *
+ * other nodes gather all slot values at once from the low nodeid.
+ * they set need_slots=1, and ignore the rf_our_slot returned from each
+ * rcom_config. they use the rf_num_slots returned from the low
+ * node's rcom_config.
+ */
+
+int dlm_rcom_status(struct dlm_ls *ls, int nodeid, uint32_t status_flags)
{
struct dlm_rcom *rc;
struct dlm_mhandle *mh;
@@ -141,10 +155,13 @@ int dlm_rcom_status(struct dlm_ls *ls, int nodeid)
goto out;
}
- error = create_rcom(ls, nodeid, DLM_RCOM_STATUS, 0, &rc, &mh);
+ error = create_rcom(ls, nodeid, DLM_RCOM_STATUS,
+ sizeof(struct rcom_status), &rc, &mh);
if (error)
goto out;
+ set_rcom_status(ls, (struct rcom_status *)rc->rc_buf, status_flags);
+
allow_sync_reply(ls, &rc->rc_id);
memset(ls->ls_recover_buf, 0, dlm_config.ci_buffer_size);
@@ -161,8 +178,11 @@ int dlm_rcom_status(struct dlm_ls *ls, int nodeid)
/* we pretend the remote lockspace exists with 0 status */
log_debug(ls, "remote node %d not ready", nodeid);
rc->rc_result = 0;
- } else
- error = check_config(ls, rc, nodeid);
+ error = 0;
+ } else {
+ error = check_rcom_config(ls, rc, nodeid);
+ }
+
/* the caller looks at rc_result for the remote recovery status */
out:
return error;
@@ -172,17 +192,60 @@ static void receive_rcom_status(struct dlm_ls *ls, struct dlm_rcom *rc_in)
{
struct dlm_rcom *rc;
struct dlm_mhandle *mh;
- int error, nodeid = rc_in->rc_header.h_nodeid;
+ struct rcom_status *rs;
+ uint32_t status;
+ int nodeid = rc_in->rc_header.h_nodeid;
+ int len = sizeof(struct rcom_config);
+ int num_slots = 0;
+ int error;
+
+ if (!dlm_slots_version(&rc_in->rc_header)) {
+ status = dlm_recover_status(ls);
+ goto do_create;
+ }
+
+ rs = (struct rcom_status *)rc_in->rc_buf;
+ if (!(rs->rs_flags & DLM_RSF_NEED_SLOTS)) {
+ status = dlm_recover_status(ls);
+ goto do_create;
+ }
+
+ spin_lock(&ls->ls_recover_lock);
+ status = ls->ls_recover_status;
+ num_slots = ls->ls_num_slots;
+ spin_unlock(&ls->ls_recover_lock);
+ len += num_slots * sizeof(struct rcom_slot);
+
+ do_create:
error = create_rcom(ls, nodeid, DLM_RCOM_STATUS_REPLY,
- sizeof(struct rcom_config), &rc, &mh);
+ len, &rc, &mh);
if (error)
return;
+
rc->rc_id = rc_in->rc_id;
rc->rc_seq_reply = rc_in->rc_seq;
- rc->rc_result = dlm_recover_status(ls);
- make_config(ls, (struct rcom_config *) rc->rc_buf);
+ rc->rc_result = status;
+ set_rcom_config(ls, (struct rcom_config *)rc->rc_buf, num_slots);
+
+ if (!num_slots)
+ goto do_send;
+
+ spin_lock(&ls->ls_recover_lock);
+ if (ls->ls_num_slots != num_slots) {
+ spin_unlock(&ls->ls_recover_lock);
+ log_debug(ls, "receive_rcom_status num_slots %d to %d",
+ num_slots, ls->ls_num_slots);
+ rc->rc_result = 0;
+ set_rcom_config(ls, (struct rcom_config *)rc->rc_buf, 0);
+ goto do_send;
+ }
+
+ dlm_slots_copy_out(ls, rc);
+ spin_unlock(&ls->ls_recover_lock);
+
+ do_send:
send_rcom(ls, mh, rc);
}
@@ -210,19 +273,9 @@ int dlm_rcom_names(struct dlm_ls *ls, int nodeid, char *last_name, int last_len)
struct dlm_rcom *rc;
struct dlm_mhandle *mh;
int error = 0;
- int max_size = dlm_config.ci_buffer_size - sizeof(struct dlm_rcom);
ls->ls_recover_nodeid = nodeid;
- if (nodeid == dlm_our_nodeid()) {
- ls->ls_recover_buf->rc_header.h_length =
- dlm_config.ci_buffer_size;
- dlm_copy_master_names(ls, last_name, last_len,
- ls->ls_recover_buf->rc_buf,
- max_size, nodeid);
- goto out;
- }
-
error = create_rcom(ls, nodeid, DLM_RCOM_NAMES, last_len, &rc, &mh);
if (error)
goto out;
@@ -272,7 +325,26 @@ int dlm_send_rcom_lookup(struct dlm_rsb *r, int dir_nodeid)
if (error)
goto out;
memcpy(rc->rc_buf, r->res_name, r->res_length);
- rc->rc_id = (unsigned long) r;
+ rc->rc_id = (unsigned long) r->res_id;
+
+ send_rcom(ls, mh, rc);
+ out:
+ return error;
+}
+
+int dlm_send_rcom_lookup_dump(struct dlm_rsb *r, int to_nodeid)
+{
+ struct dlm_rcom *rc;
+ struct dlm_mhandle *mh;
+ struct dlm_ls *ls = r->res_ls;
+ int error;
+
+ error = create_rcom(ls, to_nodeid, DLM_RCOM_LOOKUP, r->res_length,
+ &rc, &mh);
+ if (error)
+ goto out;
+ memcpy(rc->rc_buf, r->res_name, r->res_length);
+ rc->rc_id = 0xFFFFFFFF;
send_rcom(ls, mh, rc);
out:
@@ -290,7 +362,14 @@ static void receive_rcom_lookup(struct dlm_ls *ls, struct dlm_rcom *rc_in)
if (error)
return;
- error = dlm_dir_lookup(ls, nodeid, rc_in->rc_buf, len, &ret_nodeid);
+ if (rc_in->rc_id == 0xFFFFFFFF) {
+ log_error(ls, "receive_rcom_lookup dump from %d", nodeid);
+ dlm_dump_rsb_name(ls, rc_in->rc_buf, len);
+ return;
+ }
+
+ error = dlm_master_lookup(ls, nodeid, rc_in->rc_buf, len,
+ DLM_LU_RECOVER_MASTER, &ret_nodeid, NULL);
if (error)
ret_nodeid = error;
rc->rc_result = ret_nodeid;
@@ -321,9 +400,9 @@ static void pack_rcom_lock(struct dlm_rsb *r, struct dlm_lkb *lkb,
rl->rl_wait_type = cpu_to_le16(lkb->lkb_wait_type);
if (lkb->lkb_bastfn)
- rl->rl_asts |= AST_BAST;
+ rl->rl_asts |= DLM_CB_BAST;
if (lkb->lkb_astfn)
- rl->rl_asts |= AST_COMP;
+ rl->rl_asts |= DLM_CB_CAST;
rl->rl_namelen = cpu_to_le16(r->res_length);
memcpy(rl->rl_name, r->res_name, r->res_length);
@@ -421,46 +500,102 @@ int dlm_send_ls_not_ready(int nodeid, struct dlm_rcom *rc_in)
return 0;
}
-static int is_old_reply(struct dlm_ls *ls, struct dlm_rcom *rc)
+/*
+ * Ignore messages for stage Y before we set
+ * recover_status bit for stage X:
+ *
+ * recover_status = 0
+ *
+ * dlm_recover_members()
+ * - send nothing
+ * - recv nothing
+ * - ignore NAMES, NAMES_REPLY
+ * - ignore LOOKUP, LOOKUP_REPLY
+ * - ignore LOCK, LOCK_REPLY
+ *
+ * recover_status |= NODES
+ *
+ * dlm_recover_members_wait()
+ *
+ * dlm_recover_directory()
+ * - send NAMES
+ * - recv NAMES_REPLY
+ * - ignore LOOKUP, LOOKUP_REPLY
+ * - ignore LOCK, LOCK_REPLY
+ *
+ * recover_status |= DIR
+ *
+ * dlm_recover_directory_wait()
+ *
+ * dlm_recover_masters()
+ * - send LOOKUP
+ * - recv LOOKUP_REPLY
+ *
+ * dlm_recover_locks()
+ * - send LOCKS
+ * - recv LOCKS_REPLY
+ *
+ * recover_status |= LOCKS
+ *
+ * dlm_recover_locks_wait()
+ *
+ * recover_status |= DONE
+ */
+
+/* Called by dlm_recv; corresponds to dlm_receive_message() but special
+ recovery-only comms are sent through here. */
+
+void dlm_receive_rcom(struct dlm_ls *ls, struct dlm_rcom *rc, int nodeid)
{
+ int lock_size = sizeof(struct dlm_rcom) + sizeof(struct rcom_lock);
+ int stop, reply = 0, names = 0, lookup = 0, lock = 0;
+ uint32_t status;
uint64_t seq;
- int rv = 0;
switch (rc->rc_type) {
case DLM_RCOM_STATUS_REPLY:
+ reply = 1;
+ break;
+ case DLM_RCOM_NAMES:
+ names = 1;
+ break;
case DLM_RCOM_NAMES_REPLY:
+ names = 1;
+ reply = 1;
+ break;
+ case DLM_RCOM_LOOKUP:
+ lookup = 1;
+ break;
case DLM_RCOM_LOOKUP_REPLY:
+ lookup = 1;
+ reply = 1;
+ break;
+ case DLM_RCOM_LOCK:
+ lock = 1;
+ break;
case DLM_RCOM_LOCK_REPLY:
- spin_lock(&ls->ls_recover_lock);
- seq = ls->ls_recover_seq;
- spin_unlock(&ls->ls_recover_lock);
- if (rc->rc_seq_reply != seq) {
- log_debug(ls, "ignoring old reply %x from %d "
- "seq_reply %llx expect %llx",
- rc->rc_type, rc->rc_header.h_nodeid,
- (unsigned long long)rc->rc_seq_reply,
- (unsigned long long)seq);
- rv = 1;
- }
- }
- return rv;
-}
+ lock = 1;
+ reply = 1;
+ break;
+ };
-/* Called by dlm_recv; corresponds to dlm_receive_message() but special
- recovery-only comms are sent through here. */
+ spin_lock(&ls->ls_recover_lock);
+ status = ls->ls_recover_status;
+ stop = test_bit(LSFL_RECOVER_STOP, &ls->ls_flags);
+ seq = ls->ls_recover_seq;
+ spin_unlock(&ls->ls_recover_lock);
-void dlm_receive_rcom(struct dlm_ls *ls, struct dlm_rcom *rc, int nodeid)
-{
- int lock_size = sizeof(struct dlm_rcom) + sizeof(struct rcom_lock);
+ if (stop && (rc->rc_type != DLM_RCOM_STATUS))
+ goto ignore;
- if (dlm_recovery_stopped(ls) && (rc->rc_type != DLM_RCOM_STATUS)) {
- log_debug(ls, "ignoring recovery message %x from %d",
- rc->rc_type, nodeid);
- goto out;
- }
+ if (reply && (rc->rc_seq_reply != seq))
+ goto ignore;
- if (is_old_reply(ls, rc))
- goto out;
+ if (!(status & DLM_RS_NODES) && (names || lookup || lock))
+ goto ignore;
+
+ if (!(status & DLM_RS_DIR) && (lookup || lock))
+ goto ignore;
switch (rc->rc_type) {
case DLM_RCOM_STATUS:
@@ -502,10 +637,20 @@ void dlm_receive_rcom(struct dlm_ls *ls, struct dlm_rcom *rc, int nodeid)
default:
log_error(ls, "receive_rcom bad type %d", rc->rc_type);
}
-out:
+ return;
+
+ignore:
+ log_limit(ls, "dlm_receive_rcom ignore msg %d "
+ "from %d %llu %llu recover seq %llu sts %x gen %u",
+ rc->rc_type,
+ nodeid,
+ (unsigned long long)rc->rc_seq,
+ (unsigned long long)rc->rc_seq_reply,
+ (unsigned long long)seq,
+ status, ls->ls_generation);
return;
Eshort:
- log_error(ls, "recovery message %x from %d is too short",
- rc->rc_type, nodeid);
+ log_error(ls, "recovery message %d from %d is too short",
+ rc->rc_type, nodeid);
}
diff --git a/fs/dlm/rcom.h b/fs/dlm/rcom.h
index b09abd29ba3..f8e243463c1 100644
--- a/fs/dlm/rcom.h
+++ b/fs/dlm/rcom.h
@@ -14,9 +14,10 @@
#ifndef __RCOM_DOT_H__
#define __RCOM_DOT_H__
-int dlm_rcom_status(struct dlm_ls *ls, int nodeid);
+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_lookup_dump(struct dlm_rsb *r, int to_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);
diff --git a/fs/dlm/recover.c b/fs/dlm/recover.c
index eda43f36261..eaea789bf97 100644
--- a/fs/dlm/recover.c
+++ b/fs/dlm/recover.c
@@ -36,30 +36,23 @@
* (LS_RECOVERY_STOP set due to failure of a node in ls_nodes). When another
* function thinks it could have completed the waited-on task, they should wake
* up ls_wait_general to get an immediate response rather than waiting for the
- * timer to detect the result. A timer wakes us up periodically while waiting
- * to see if we should abort due to a node failure. This should only be called
- * by the dlm_recoverd thread.
+ * timeout. This uses a timeout so it can check periodically if the wait
+ * should abort due to node failure (which doesn't cause a wake_up).
+ * This should only be called by the dlm_recoverd thread.
*/
-static void dlm_wait_timer_fn(unsigned long data)
-{
- struct dlm_ls *ls = (struct dlm_ls *) data;
- mod_timer(&ls->ls_timer, jiffies + (dlm_config.ci_recover_timer * HZ));
- wake_up(&ls->ls_wait_general);
-}
-
int dlm_wait_function(struct dlm_ls *ls, int (*testfn) (struct dlm_ls *ls))
{
int error = 0;
+ int rv;
- init_timer(&ls->ls_timer);
- ls->ls_timer.function = dlm_wait_timer_fn;
- ls->ls_timer.data = (long) ls;
- ls->ls_timer.expires = jiffies + (dlm_config.ci_recover_timer * HZ);
- add_timer(&ls->ls_timer);
-
- wait_event(ls->ls_wait_general, testfn(ls) || dlm_recovery_stopped(ls));
- del_timer_sync(&ls->ls_timer);
+ while (1) {
+ rv = wait_event_timeout(ls->ls_wait_general,
+ testfn(ls) || dlm_recovery_stopped(ls),
+ dlm_config.ci_recover_timer * HZ);
+ if (rv)
+ break;
+ }
if (dlm_recovery_stopped(ls)) {
log_debug(ls, "dlm_wait_function aborted");
@@ -85,14 +78,20 @@ uint32_t dlm_recover_status(struct dlm_ls *ls)
return status;
}
+static void _set_recover_status(struct dlm_ls *ls, uint32_t status)
+{
+ ls->ls_recover_status |= status;
+}
+
void dlm_set_recover_status(struct dlm_ls *ls, uint32_t status)
{
spin_lock(&ls->ls_recover_lock);
- ls->ls_recover_status |= status;
+ _set_recover_status(ls, status);
spin_unlock(&ls->ls_recover_lock);
}
-static int wait_status_all(struct dlm_ls *ls, uint32_t wait_status)
+static int wait_status_all(struct dlm_ls *ls, uint32_t wait_status,
+ int save_slots)
{
struct dlm_rcom *rc = ls->ls_recover_buf;
struct dlm_member *memb;
@@ -106,10 +105,13 @@ static int wait_status_all(struct dlm_ls *ls, uint32_t wait_status)
goto out;
}
- error = dlm_rcom_status(ls, memb->nodeid);
+ error = dlm_rcom_status(ls, memb->nodeid, 0);
if (error)
goto out;
+ if (save_slots)
+ dlm_slot_save(ls, rc, memb);
+
if (rc->rc_result & wait_status)
break;
if (delay < 1000)
@@ -121,7 +123,8 @@ static int wait_status_all(struct dlm_ls *ls, uint32_t wait_status)
return error;
}
-static int wait_status_low(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)
{
struct dlm_rcom *rc = ls->ls_recover_buf;
int error = 0, delay = 0, nodeid = ls->ls_low_nodeid;
@@ -132,7 +135,7 @@ static int wait_status_low(struct dlm_ls *ls, uint32_t wait_status)
goto out;
}
- error = dlm_rcom_status(ls, nodeid);
+ error = dlm_rcom_status(ls, nodeid, status_flags);
if (error)
break;
@@ -152,18 +155,56 @@ static int wait_status(struct dlm_ls *ls, uint32_t status)
int error;
if (ls->ls_low_nodeid == dlm_our_nodeid()) {
- error = wait_status_all(ls, status);
+ error = wait_status_all(ls, status, 0);
if (!error)
dlm_set_recover_status(ls, status_all);
} else
- error = wait_status_low(ls, status_all);
+ error = wait_status_low(ls, status_all, 0);
return error;
}
int dlm_recover_members_wait(struct dlm_ls *ls)
{
- return wait_status(ls, DLM_RS_NODES);
+ struct dlm_member *memb;
+ struct dlm_slot *slots;
+ int num_slots, slots_size;
+ int error, rv;
+ uint32_t gen;
+
+ list_for_each_entry(memb, &ls->ls_nodes, list) {
+ memb->slot = -1;
+ memb->generation = 0;
+ }
+
+ if (ls->ls_low_nodeid == dlm_our_nodeid()) {
+ error = wait_status_all(ls, DLM_RS_NODES, 1);
+ if (error)
+ goto out;
+
+ /* slots array is sparse, slots_size may be > num_slots */
+
+ rv = dlm_slots_assign(ls, &num_slots, &slots_size, &slots, &gen);
+ if (!rv) {
+ spin_lock(&ls->ls_recover_lock);
+ _set_recover_status(ls, DLM_RS_NODES_ALL);
+ ls->ls_num_slots = num_slots;
+ ls->ls_slots_size = slots_size;
+ ls->ls_slots = slots;
+ ls->ls_generation = gen;
+ spin_unlock(&ls->ls_recover_lock);
+ } else {
+ dlm_set_recover_status(ls, DLM_RS_NODES_ALL);
+ }
+ } else {
+ error = wait_status_low(ls, DLM_RS_NODES_ALL, DLM_RSF_NEED_SLOTS);
+ if (error)
+ goto out;
+
+ dlm_slots_copy_in(ls);
+ }
+ out:
+ return error;
}
int dlm_recover_directory_wait(struct dlm_ls *ls)
@@ -229,32 +270,100 @@ static void recover_list_del(struct dlm_rsb *r)
dlm_put_rsb(r);
}
-static struct dlm_rsb *recover_list_find(struct dlm_ls *ls, uint64_t id)
+static void recover_list_clear(struct dlm_ls *ls)
{
- struct dlm_rsb *r = NULL;
+ struct dlm_rsb *r, *s;
spin_lock(&ls->ls_recover_list_lock);
+ list_for_each_entry_safe(r, s, &ls->ls_recover_list, res_recover_list) {
+ list_del_init(&r->res_recover_list);
+ r->res_recover_locks_count = 0;
+ dlm_put_rsb(r);
+ ls->ls_recover_list_count--;
+ }
- list_for_each_entry(r, &ls->ls_recover_list, res_recover_list) {
- if (id == (unsigned long) r)
- goto out;
+ if (ls->ls_recover_list_count != 0) {
+ log_error(ls, "warning: recover_list_count %d",
+ ls->ls_recover_list_count);
+ ls->ls_recover_list_count = 0;
}
- r = NULL;
- out:
spin_unlock(&ls->ls_recover_list_lock);
+}
+
+static int recover_idr_empty(struct dlm_ls *ls)
+{
+ int empty = 1;
+
+ spin_lock(&ls->ls_recover_idr_lock);
+ if (ls->ls_recover_list_count)
+ empty = 0;
+ spin_unlock(&ls->ls_recover_idr_lock);
+
+ return empty;
+}
+
+static int recover_idr_add(struct dlm_rsb *r)
+{
+ struct dlm_ls *ls = r->res_ls;
+ int rv;
+
+ idr_preload(GFP_NOFS);
+ spin_lock(&ls->ls_recover_idr_lock);
+ if (r->res_id) {
+ rv = -1;
+ goto out_unlock;
+ }
+ rv = idr_alloc(&ls->ls_recover_idr, r, 1, 0, GFP_NOWAIT);
+ if (rv < 0)
+ goto out_unlock;
+
+ r->res_id = rv;
+ ls->ls_recover_list_count++;
+ dlm_hold_rsb(r);
+ rv = 0;
+out_unlock:
+ spin_unlock(&ls->ls_recover_idr_lock);
+ idr_preload_end();
+ return rv;
+}
+
+static void recover_idr_del(struct dlm_rsb *r)
+{
+ struct dlm_ls *ls = r->res_ls;
+
+ spin_lock(&ls->ls_recover_idr_lock);
+ idr_remove(&ls->ls_recover_idr, r->res_id);
+ r->res_id = 0;
+ ls->ls_recover_list_count--;
+ spin_unlock(&ls->ls_recover_idr_lock);
+
+ dlm_put_rsb(r);
+}
+
+static struct dlm_rsb *recover_idr_find(struct dlm_ls *ls, uint64_t id)
+{
+ struct dlm_rsb *r;
+
+ spin_lock(&ls->ls_recover_idr_lock);
+ r = idr_find(&ls->ls_recover_idr, (int)id);
+ spin_unlock(&ls->ls_recover_idr_lock);
return r;
}
-static void recover_list_clear(struct dlm_ls *ls)
+static void recover_idr_clear(struct dlm_ls *ls)
{
- struct dlm_rsb *r, *s;
+ struct dlm_rsb *r;
+ int id;
- spin_lock(&ls->ls_recover_list_lock);
- list_for_each_entry_safe(r, s, &ls->ls_recover_list, res_recover_list) {
- list_del_init(&r->res_recover_list);
+ spin_lock(&ls->ls_recover_idr_lock);
+
+ idr_for_each_entry(&ls->ls_recover_idr, r, id) {
+ idr_remove(&ls->ls_recover_idr, id);
+ r->res_id = 0;
r->res_recover_locks_count = 0;
- dlm_put_rsb(r);
ls->ls_recover_list_count--;
+
+ dlm_put_rsb(r);
}
if (ls->ls_recover_list_count != 0) {
@@ -262,7 +371,7 @@ static void recover_list_clear(struct dlm_ls *ls)
ls->ls_recover_list_count);
ls->ls_recover_list_count = 0;
}
- spin_unlock(&ls->ls_recover_list_lock);
+ spin_unlock(&ls->ls_recover_idr_lock);
}
@@ -291,9 +400,12 @@ static void set_lock_master(struct list_head *queue, int nodeid)
{
struct dlm_lkb *lkb;
- list_for_each_entry(lkb, queue, lkb_statequeue)
- if (!(lkb->lkb_flags & DLM_IFL_MSTCPY))
+ list_for_each_entry(lkb, queue, lkb_statequeue) {
+ if (!(lkb->lkb_flags & DLM_IFL_MSTCPY)) {
lkb->lkb_nodeid = nodeid;
+ lkb->lkb_remid = 0;
+ }
+ }
}
static void set_master_lkbs(struct dlm_rsb *r)
@@ -304,69 +416,95 @@ static void set_master_lkbs(struct dlm_rsb *r)
}
/*
- * Propogate the new master nodeid to locks
+ * Propagate the new master nodeid to locks
* The NEW_MASTER flag tells dlm_recover_locks() which rsb's to consider.
- * The NEW_MASTER2 flag tells recover_lvb() and set_locks_purged() which
+ * The NEW_MASTER2 flag tells recover_lvb() and recover_grant() which
* rsb's to consider.
*/
-static void set_new_master(struct dlm_rsb *r, int nodeid)
+static void set_new_master(struct dlm_rsb *r)
{
- lock_rsb(r);
- r->res_nodeid = nodeid;
set_master_lkbs(r);
rsb_set_flag(r, RSB_NEW_MASTER);
rsb_set_flag(r, RSB_NEW_MASTER2);
- unlock_rsb(r);
}
/*
* We do async lookups on rsb's that need new masters. The rsb's
* waiting for a lookup reply are kept on the recover_list.
+ *
+ * Another node recovering the master may have sent us a rcom lookup,
+ * and our dlm_master_lookup() set it as the new master, along with
+ * NEW_MASTER so that we'll recover it here (this implies dir_nodeid
+ * equals our_nodeid below).
*/
-static int recover_master(struct dlm_rsb *r)
+static int recover_master(struct dlm_rsb *r, unsigned int *count)
{
struct dlm_ls *ls = r->res_ls;
- int error, dir_nodeid, ret_nodeid, our_nodeid = dlm_our_nodeid();
+ int our_nodeid, dir_nodeid;
+ int is_removed = 0;
+ int error;
+ if (is_master(r))
+ return 0;
+
+ is_removed = dlm_is_removed(ls, r->res_nodeid);
+
+ if (!is_removed && !rsb_flag(r, RSB_NEW_MASTER))
+ return 0;
+
+ our_nodeid = dlm_our_nodeid();
dir_nodeid = dlm_dir_nodeid(r);
if (dir_nodeid == our_nodeid) {
- error = dlm_dir_lookup(ls, our_nodeid, r->res_name,
- r->res_length, &ret_nodeid);
- if (error)
- log_error(ls, "recover dir lookup error %d", error);
+ if (is_removed) {
+ r->res_master_nodeid = our_nodeid;
+ r->res_nodeid = 0;
+ }
- if (ret_nodeid == our_nodeid)
- ret_nodeid = 0;
- set_new_master(r, ret_nodeid);
+ /* set master of lkbs to ourself when is_removed, or to
+ another new master which we set along with NEW_MASTER
+ in dlm_master_lookup */
+ set_new_master(r);
+ error = 0;
} else {
- recover_list_add(r);
+ recover_idr_add(r);
error = dlm_send_rcom_lookup(r, dir_nodeid);
}
+ (*count)++;
return error;
}
/*
- * When not using a directory, most resource names will hash to a new static
- * master nodeid and the resource will need to be remastered.
+ * All MSTCPY locks are purged and rebuilt, even if the master stayed the same.
+ * This is necessary because recovery can be started, aborted and restarted,
+ * causing the master nodeid to briefly change during the aborted recovery, and
+ * change back to the original value in the second recovery. The MSTCPY locks
+ * may or may not have been purged during the aborted recovery. Another node
+ * with an outstanding request in waiters list and a request reply saved in the
+ * requestqueue, cannot know whether it should ignore the reply and resend the
+ * request, or accept the reply and complete the request. It must do the
+ * former if the remote node purged MSTCPY locks, and it must do the later if
+ * the remote node did not. This is solved by always purging MSTCPY locks, in
+ * which case, the request reply would always be ignored and the request
+ * resent.
*/
-static int recover_master_static(struct dlm_rsb *r)
+static int recover_master_static(struct dlm_rsb *r, unsigned int *count)
{
- int master = dlm_dir_nodeid(r);
+ int dir_nodeid = dlm_dir_nodeid(r);
+ int new_master = dir_nodeid;
- if (master == dlm_our_nodeid())
- master = 0;
+ if (dir_nodeid == dlm_our_nodeid())
+ new_master = 0;
- if (r->res_nodeid != master) {
- if (is_master(r))
- dlm_purge_mstcpy_locks(r);
- set_new_master(r, master);
- return 1;
- }
+ dlm_purge_mstcpy_locks(r);
+ r->res_master_nodeid = dir_nodeid;
+ r->res_nodeid = new_master;
+ set_new_master(r);
+ (*count)++;
return 0;
}
@@ -383,9 +521,12 @@ static int recover_master_static(struct dlm_rsb *r)
int dlm_recover_masters(struct dlm_ls *ls)
{
struct dlm_rsb *r;
- int error = 0, count = 0;
+ unsigned int total = 0;
+ unsigned int count = 0;
+ int nodir = dlm_no_directory(ls);
+ int error;
- log_debug(ls, "dlm_recover_masters");
+ log_rinfo(ls, "dlm_recover_masters");
down_read(&ls->ls_root_sem);
list_for_each_entry(r, &ls->ls_root_list, res_root_list) {
@@ -395,48 +536,58 @@ int dlm_recover_masters(struct dlm_ls *ls)
goto out;
}
- if (dlm_no_directory(ls))
- count += recover_master_static(r);
- else if (!is_master(r) &&
- (dlm_is_removed(ls, r->res_nodeid) ||
- rsb_flag(r, RSB_NEW_MASTER))) {
- recover_master(r);
- count++;
- }
+ lock_rsb(r);
+ if (nodir)
+ error = recover_master_static(r, &count);
+ else
+ error = recover_master(r, &count);
+ unlock_rsb(r);
+ cond_resched();
+ total++;
- schedule();
+ if (error) {
+ up_read(&ls->ls_root_sem);
+ goto out;
+ }
}
up_read(&ls->ls_root_sem);
- log_debug(ls, "dlm_recover_masters %d resources", count);
+ log_rinfo(ls, "dlm_recover_masters %u of %u", count, total);
- error = dlm_wait_function(ls, &recover_list_empty);
+ error = dlm_wait_function(ls, &recover_idr_empty);
out:
if (error)
- recover_list_clear(ls);
+ recover_idr_clear(ls);
return error;
}
int dlm_recover_master_reply(struct dlm_ls *ls, struct dlm_rcom *rc)
{
struct dlm_rsb *r;
- int nodeid;
+ int ret_nodeid, new_master;
- r = recover_list_find(ls, rc->rc_id);
+ r = recover_idr_find(ls, rc->rc_id);
if (!r) {
log_error(ls, "dlm_recover_master_reply no id %llx",
(unsigned long long)rc->rc_id);
goto out;
}
- nodeid = rc->rc_result;
- if (nodeid == dlm_our_nodeid())
- nodeid = 0;
+ ret_nodeid = rc->rc_result;
- set_new_master(r, nodeid);
- recover_list_del(r);
+ if (ret_nodeid == dlm_our_nodeid())
+ new_master = 0;
+ else
+ new_master = ret_nodeid;
- if (recover_list_empty(ls))
+ lock_rsb(r);
+ r->res_master_nodeid = ret_nodeid;
+ r->res_nodeid = new_master;
+ set_new_master(r);
+ unlock_rsb(r);
+ recover_idr_del(r);
+
+ if (recover_idr_empty(ls))
wake_up(&ls->ls_wait_general);
out:
return 0;
@@ -508,8 +659,6 @@ int dlm_recover_locks(struct dlm_ls *ls)
struct dlm_rsb *r;
int error, count = 0;
- log_debug(ls, "dlm_recover_locks");
-
down_read(&ls->ls_root_sem);
list_for_each_entry(r, &ls->ls_root_list, res_root_list) {
if (is_master(r)) {
@@ -536,14 +685,12 @@ int dlm_recover_locks(struct dlm_ls *ls)
}
up_read(&ls->ls_root_sem);
- log_debug(ls, "dlm_recover_locks %d locks", count);
+ log_rinfo(ls, "dlm_recover_locks %d out", count);
error = dlm_wait_function(ls, &recover_list_empty);
out:
if (error)
recover_list_clear(ls);
- else
- dlm_set_recover_status(ls, DLM_RS_LOCKS);
return error;
}
@@ -566,8 +713,14 @@ void dlm_recovered_lock(struct dlm_rsb *r)
* the VALNOTVALID flag if necessary, and determining the correct lvb contents
* based on the lvb's of the locks held on the rsb.
*
- * RSB_VALNOTVALID is set if there are only NL/CR locks on the rsb. If it
- * was already set prior to recovery, it's not cleared, regardless of locks.
+ * RSB_VALNOTVALID is set in two cases:
+ *
+ * 1. we are master, but not new, and we purged an EX/PW lock held by a
+ * failed node (in dlm_recover_purge which set RSB_RECOVER_LVB_INVAL)
+ *
+ * 2. we are a new master, and there are only NL/CR locks left.
+ * (We could probably improve this by only invaliding in this way when
+ * the previous master left uncleanly. VMS docs mention that.)
*
* The LVB contents are only considered for changing when this is a new master
* of the rsb (NEW_MASTER2). Then, the rsb's lvb is taken from any lkb with
@@ -583,6 +736,19 @@ static void recover_lvb(struct dlm_rsb *r)
int big_lock_exists = 0;
int lvblen = r->res_ls->ls_lvblen;
+ if (!rsb_flag(r, RSB_NEW_MASTER2) &&
+ rsb_flag(r, RSB_RECOVER_LVB_INVAL)) {
+ /* case 1 above */
+ rsb_set_flag(r, RSB_VALNOTVALID);
+ return;
+ }
+
+ if (!rsb_flag(r, RSB_NEW_MASTER2))
+ return;
+
+ /* we are the new master, so figure out if VALNOTVALID should
+ be set, and set the rsb lvb from the best lkb available. */
+
list_for_each_entry(lkb, &r->res_grantqueue, lkb_statequeue) {
if (!(lkb->lkb_exflags & DLM_LKF_VALBLK))
continue;
@@ -621,13 +787,10 @@ static void recover_lvb(struct dlm_rsb *r)
if (!lock_lvb_exists)
goto out;
+ /* lvb is invalidated if only NL/CR locks remain */
if (!big_lock_exists)
rsb_set_flag(r, RSB_VALNOTVALID);
- /* don't mess with the lvb unless we're the new master */
- if (!rsb_flag(r, RSB_NEW_MASTER2))
- goto out;
-
if (!r->res_lvbptr) {
r->res_lvbptr = dlm_allocate_lvb(r->res_ls);
if (!r->res_lvbptr)
@@ -653,6 +816,7 @@ static void recover_lvb(struct dlm_rsb *r)
static void recover_conversion(struct dlm_rsb *r)
{
+ struct dlm_ls *ls = r->res_ls;
struct dlm_lkb *lkb;
int grmode = -1;
@@ -667,29 +831,32 @@ static void recover_conversion(struct dlm_rsb *r)
list_for_each_entry(lkb, &r->res_convertqueue, lkb_statequeue) {
if (lkb->lkb_grmode != DLM_LOCK_IV)
continue;
- if (grmode == -1)
+ if (grmode == -1) {
+ log_debug(ls, "recover_conversion %x set gr to rq %d",
+ lkb->lkb_id, lkb->lkb_rqmode);
lkb->lkb_grmode = lkb->lkb_rqmode;
- else
+ } else {
+ log_debug(ls, "recover_conversion %x set gr %d",
+ lkb->lkb_id, grmode);
lkb->lkb_grmode = grmode;
+ }
}
}
/* We've become the new master for this rsb and waiting/converting locks may
- need to be granted in dlm_grant_after_purge() due to locks that may have
+ need to be granted in dlm_recover_grant() due to locks that may have
existed from a removed node. */
-static void set_locks_purged(struct dlm_rsb *r)
+static void recover_grant(struct dlm_rsb *r)
{
if (!list_empty(&r->res_waitqueue) || !list_empty(&r->res_convertqueue))
- rsb_set_flag(r, RSB_LOCKS_PURGED);
+ rsb_set_flag(r, RSB_RECOVER_GRANT);
}
void dlm_recover_rsbs(struct dlm_ls *ls)
{
struct dlm_rsb *r;
- int count = 0;
-
- log_debug(ls, "dlm_recover_rsbs");
+ unsigned int count = 0;
down_read(&ls->ls_root_sem);
list_for_each_entry(r, &ls->ls_root_list, res_root_list) {
@@ -697,24 +864,33 @@ void dlm_recover_rsbs(struct dlm_ls *ls)
if (is_master(r)) {
if (rsb_flag(r, RSB_RECOVER_CONVERT))
recover_conversion(r);
- if (rsb_flag(r, RSB_NEW_MASTER2))
- set_locks_purged(r);
+
+ /* recover lvb before granting locks so the updated
+ lvb/VALNOTVALID is presented in the completion */
recover_lvb(r);
+
+ if (rsb_flag(r, RSB_NEW_MASTER2))
+ recover_grant(r);
count++;
+ } else {
+ rsb_clear_flag(r, RSB_VALNOTVALID);
}
rsb_clear_flag(r, RSB_RECOVER_CONVERT);
+ rsb_clear_flag(r, RSB_RECOVER_LVB_INVAL);
rsb_clear_flag(r, RSB_NEW_MASTER2);
unlock_rsb(r);
}
up_read(&ls->ls_root_sem);
- log_debug(ls, "dlm_recover_rsbs %d rsbs", count);
+ if (count)
+ log_rinfo(ls, "dlm_recover_rsbs %d done", count);
}
/* Create a single list of all root rsb's to be used during recovery */
int dlm_create_root_list(struct dlm_ls *ls)
{
+ struct rb_node *n;
struct dlm_rsb *r;
int i, error = 0;
@@ -727,24 +903,14 @@ int dlm_create_root_list(struct dlm_ls *ls)
for (i = 0; i < ls->ls_rsbtbl_size; i++) {
spin_lock(&ls->ls_rsbtbl[i].lock);
- list_for_each_entry(r, &ls->ls_rsbtbl[i].list, res_hashchain) {
+ for (n = rb_first(&ls->ls_rsbtbl[i].keep); n; n = rb_next(n)) {
+ r = rb_entry(n, struct dlm_rsb, res_hashnode);
list_add(&r->res_root_list, &ls->ls_root_list);
dlm_hold_rsb(r);
}
- /* If we're using a directory, add tossed rsbs to the root
- list; they'll have entries created in the new directory,
- but no other recovery steps should do anything with them. */
-
- if (dlm_no_directory(ls)) {
- spin_unlock(&ls->ls_rsbtbl[i].lock);
- continue;
- }
-
- list_for_each_entry(r, &ls->ls_rsbtbl[i].toss, res_hashchain) {
- list_add(&r->res_root_list, &ls->ls_root_list);
- dlm_hold_rsb(r);
- }
+ if (!RB_EMPTY_ROOT(&ls->ls_rsbtbl[i].toss))
+ log_error(ls, "dlm_create_root_list toss not empty");
spin_unlock(&ls->ls_rsbtbl[i].lock);
}
out:
@@ -764,26 +930,26 @@ void dlm_release_root_list(struct dlm_ls *ls)
up_write(&ls->ls_root_sem);
}
-/* If not using a directory, clear the entire toss list, there's no benefit to
- caching the master value since it's fixed. If we are using a dir, keep the
- rsb's we're the master of. Recovery will add them to the root list and from
- there they'll be entered in the rebuilt directory. */
-
-void dlm_clear_toss_list(struct dlm_ls *ls)
+void dlm_clear_toss(struct dlm_ls *ls)
{
- struct dlm_rsb *r, *safe;
+ struct rb_node *n, *next;
+ struct dlm_rsb *r;
+ unsigned int count = 0;
int i;
for (i = 0; i < ls->ls_rsbtbl_size; i++) {
spin_lock(&ls->ls_rsbtbl[i].lock);
- list_for_each_entry_safe(r, safe, &ls->ls_rsbtbl[i].toss,
- res_hashchain) {
- if (dlm_no_directory(ls) || !is_master(r)) {
- list_del(&r->res_hashchain);
- dlm_free_rsb(r);
- }
+ for (n = rb_first(&ls->ls_rsbtbl[i].toss); n; n = next) {
+ next = rb_next(n);
+ r = rb_entry(n, struct dlm_rsb, res_hashnode);
+ rb_erase(n, &ls->ls_rsbtbl[i].toss);
+ dlm_free_rsb(r);
+ count++;
}
spin_unlock(&ls->ls_rsbtbl[i].lock);
}
+
+ if (count)
+ log_rinfo(ls, "dlm_clear_toss %u done", count);
}
diff --git a/fs/dlm/recover.h b/fs/dlm/recover.h
index ebd0363f1e0..d8c8738c70e 100644
--- a/fs/dlm/recover.h
+++ b/fs/dlm/recover.h
@@ -27,7 +27,7 @@ int dlm_recover_locks(struct dlm_ls *ls);
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);
-void dlm_clear_toss_list(struct dlm_ls *ls);
+void dlm_clear_toss(struct dlm_ls *ls);
void dlm_recover_rsbs(struct dlm_ls *ls);
#endif /* __RECOVER_DOT_H__ */
diff --git a/fs/dlm/recoverd.c b/fs/dlm/recoverd.c
index fd677c8c3d3..6859b4bf971 100644
--- a/fs/dlm/recoverd.c
+++ b/fs/dlm/recoverd.c
@@ -2,7 +2,7 @@
*******************************************************************************
**
** Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
-** Copyright (C) 2004-2007 Red Hat, Inc. All rights reserved.
+** Copyright (C) 2004-2011 Red Hat, Inc. All rights reserved.
**
** This copyrighted material is made available to anyone wishing to use,
** modify, copy, or redistribute it subject to the terms and conditions
@@ -41,6 +41,7 @@ static int enable_locking(struct dlm_ls *ls, uint64_t seq)
set_bit(LSFL_RUNNING, &ls->ls_flags);
/* unblocks processes waiting to enter the dlm */
up_write(&ls->ls_in_recovery);
+ clear_bit(LSFL_RECOVER_LOCK, &ls->ls_flags);
error = 0;
}
spin_unlock(&ls->ls_recover_lock);
@@ -54,24 +55,13 @@ static int ls_recover(struct dlm_ls *ls, struct dlm_recover *rv)
unsigned long start;
int error, neg = 0;
- log_debug(ls, "recover %llx", (unsigned long long)rv->seq);
+ log_rinfo(ls, "dlm_recover %llu", (unsigned long long)rv->seq);
mutex_lock(&ls->ls_recoverd_active);
- /*
- * Suspending and resuming dlm_astd ensures that no lkb's from this ls
- * will be processed by dlm_astd during recovery.
- */
-
- dlm_astd_suspend();
- dlm_astd_resume();
-
- /*
- * Free non-master tossed rsb's. Master rsb's are kept on toss
- * list and put on root list to be included in resdir recovery.
- */
+ dlm_callback_suspend(ls);
- dlm_clear_toss_list(ls);
+ dlm_clear_toss(ls);
/*
* This list of root rsb's will be the basis of most of the recovery
@@ -82,14 +72,28 @@ static int ls_recover(struct dlm_ls *ls, struct dlm_recover *rv)
/*
* Add or remove nodes from the lockspace's ls_nodes list.
- * Also waits for all nodes to complete dlm_recover_members.
*/
error = dlm_recover_members(ls, rv, &neg);
if (error) {
- log_debug(ls, "recover_members failed %d", error);
+ log_rinfo(ls, "dlm_recover_members error %d", error);
+ goto fail;
+ }
+
+ dlm_recover_dir_nodeid(ls);
+
+ ls->ls_recover_dir_sent_res = 0;
+ ls->ls_recover_dir_sent_msg = 0;
+ ls->ls_recover_locks_in = 0;
+
+ dlm_set_recover_status(ls, DLM_RS_NODES);
+
+ error = dlm_recover_members_wait(ls);
+ if (error) {
+ log_rinfo(ls, "dlm_recover_members_wait error %d", error);
goto fail;
}
+
start = jiffies;
/*
@@ -99,20 +103,21 @@ static int ls_recover(struct dlm_ls *ls, struct dlm_recover *rv)
error = dlm_recover_directory(ls);
if (error) {
- log_debug(ls, "recover_directory failed %d", error);
+ log_rinfo(ls, "dlm_recover_directory error %d", error);
goto fail;
}
- /*
- * Wait for all nodes to complete directory rebuild.
- */
+ dlm_set_recover_status(ls, DLM_RS_DIR);
error = dlm_recover_directory_wait(ls);
if (error) {
- log_debug(ls, "recover_directory_wait failed %d", error);
+ log_rinfo(ls, "dlm_recover_directory_wait error %d", error);
goto fail;
}
+ log_rinfo(ls, "dlm_recover_directory %u out %u messages",
+ ls->ls_recover_dir_sent_res, ls->ls_recover_dir_sent_msg);
+
/*
* We may have outstanding operations that are waiting for a reply from
* a failed node. Mark these to be resent after recovery. Unlock and
@@ -130,7 +135,7 @@ static int ls_recover(struct dlm_ls *ls, struct dlm_recover *rv)
* Clear lkb's for departed nodes.
*/
- dlm_purge_locks(ls);
+ dlm_recover_purge(ls);
/*
* Get new master nodeid's for rsb's that were mastered on
@@ -139,7 +144,7 @@ static int ls_recover(struct dlm_ls *ls, struct dlm_recover *rv)
error = dlm_recover_masters(ls);
if (error) {
- log_debug(ls, "recover_masters failed %d", error);
+ log_rinfo(ls, "dlm_recover_masters error %d", error);
goto fail;
}
@@ -149,16 +154,21 @@ static int ls_recover(struct dlm_ls *ls, struct dlm_recover *rv)
error = dlm_recover_locks(ls);
if (error) {
- log_debug(ls, "recover_locks failed %d", error);
+ log_rinfo(ls, "dlm_recover_locks error %d", error);
goto fail;
}
+ dlm_set_recover_status(ls, DLM_RS_LOCKS);
+
error = dlm_recover_locks_wait(ls);
if (error) {
- log_debug(ls, "recover_locks_wait failed %d", error);
+ log_rinfo(ls, "dlm_recover_locks_wait error %d", error);
goto fail;
}
+ log_rinfo(ls, "dlm_recover_locks %u in",
+ ls->ls_recover_locks_in);
+
/*
* Finalize state in master rsb's now that all locks can be
* checked. This includes conversion resolution and lvb
@@ -176,7 +186,7 @@ static int ls_recover(struct dlm_ls *ls, struct dlm_recover *rv)
error = dlm_recover_locks_wait(ls);
if (error) {
- log_debug(ls, "recover_locks_wait failed %d", error);
+ log_rinfo(ls, "dlm_recover_locks_wait error %d", error);
goto fail;
}
}
@@ -192,9 +202,10 @@ static int ls_recover(struct dlm_ls *ls, struct dlm_recover *rv)
dlm_purge_requestqueue(ls);
dlm_set_recover_status(ls, DLM_RS_DONE);
+
error = dlm_recover_done_wait(ls);
if (error) {
- log_debug(ls, "recover_done_wait failed %d", error);
+ log_rinfo(ls, "dlm_recover_done_wait error %d", error);
goto fail;
}
@@ -202,38 +213,39 @@ static int ls_recover(struct dlm_ls *ls, struct dlm_recover *rv)
dlm_adjust_timeouts(ls);
+ dlm_callback_resume(ls);
+
error = enable_locking(ls, rv->seq);
if (error) {
- log_debug(ls, "enable_locking failed %d", error);
+ log_rinfo(ls, "enable_locking error %d", error);
goto fail;
}
error = dlm_process_requestqueue(ls);
if (error) {
- log_debug(ls, "process_requestqueue failed %d", error);
+ log_rinfo(ls, "dlm_process_requestqueue error %d", error);
goto fail;
}
error = dlm_recover_waiters_post(ls);
if (error) {
- log_debug(ls, "recover_waiters_post failed %d", error);
+ log_rinfo(ls, "dlm_recover_waiters_post error %d", error);
goto fail;
}
- dlm_grant_after_purge(ls);
+ dlm_recover_grant(ls);
- dlm_astd_wake();
-
- log_debug(ls, "recover %llx done: %u ms",
- (unsigned long long)rv->seq,
+ log_rinfo(ls, "dlm_recover %llu generation %u done: %u ms",
+ (unsigned long long)rv->seq, ls->ls_generation,
jiffies_to_msecs(jiffies - start));
mutex_unlock(&ls->ls_recoverd_active);
+ dlm_lsop_recover_done(ls);
return 0;
fail:
dlm_release_root_list(ls);
- log_debug(ls, "recover %llx error %d",
+ log_rinfo(ls, "dlm_recover %llu error %d",
(unsigned long long)rv->seq, error);
mutex_unlock(&ls->ls_recoverd_active);
return error;
@@ -251,13 +263,12 @@ static void do_ls_recovery(struct dlm_ls *ls)
rv = ls->ls_recover_args;
ls->ls_recover_args = NULL;
if (rv && ls->ls_recover_seq == rv->seq)
- clear_bit(LSFL_RECOVERY_STOP, &ls->ls_flags);
+ clear_bit(LSFL_RECOVER_STOP, &ls->ls_flags);
spin_unlock(&ls->ls_recover_lock);
if (rv) {
ls_recover(ls, rv);
- kfree(rv->nodeids);
- kfree(rv->new);
+ kfree(rv->nodes);
kfree(rv);
}
}
@@ -272,26 +283,34 @@ static int dlm_recoverd(void *arg)
return -1;
}
+ down_write(&ls->ls_in_recovery);
+ set_bit(LSFL_RECOVER_LOCK, &ls->ls_flags);
+ wake_up(&ls->ls_recover_lock_wait);
+
while (!kthread_should_stop()) {
set_current_state(TASK_INTERRUPTIBLE);
- if (!test_bit(LSFL_WORK, &ls->ls_flags))
+ if (!test_bit(LSFL_RECOVER_WORK, &ls->ls_flags) &&
+ !test_bit(LSFL_RECOVER_DOWN, &ls->ls_flags))
schedule();
set_current_state(TASK_RUNNING);
- if (test_and_clear_bit(LSFL_WORK, &ls->ls_flags))
+ if (test_and_clear_bit(LSFL_RECOVER_DOWN, &ls->ls_flags)) {
+ down_write(&ls->ls_in_recovery);
+ set_bit(LSFL_RECOVER_LOCK, &ls->ls_flags);
+ wake_up(&ls->ls_recover_lock_wait);
+ }
+
+ if (test_and_clear_bit(LSFL_RECOVER_WORK, &ls->ls_flags))
do_ls_recovery(ls);
}
+ if (test_bit(LSFL_RECOVER_LOCK, &ls->ls_flags))
+ up_write(&ls->ls_in_recovery);
+
dlm_put_lockspace(ls);
return 0;
}
-void dlm_recoverd_kick(struct dlm_ls *ls)
-{
- set_bit(LSFL_WORK, &ls->ls_flags);
- wake_up_process(ls->ls_recoverd_task);
-}
-
int dlm_recoverd_start(struct dlm_ls *ls)
{
struct task_struct *p;
diff --git a/fs/dlm/recoverd.h b/fs/dlm/recoverd.h
index 866657c5d69..8856079733f 100644
--- a/fs/dlm/recoverd.h
+++ b/fs/dlm/recoverd.h
@@ -14,7 +14,6 @@
#ifndef __RECOVERD_DOT_H__
#define __RECOVERD_DOT_H__
-void dlm_recoverd_kick(struct dlm_ls *ls);
void dlm_recoverd_stop(struct dlm_ls *ls);
int dlm_recoverd_start(struct dlm_ls *ls);
void dlm_recoverd_suspend(struct dlm_ls *ls);
diff --git a/fs/dlm/requestqueue.c b/fs/dlm/requestqueue.c
index a44fa22890e..1695f1b0dd4 100644
--- a/fs/dlm/requestqueue.c
+++ b/fs/dlm/requestqueue.c
@@ -19,6 +19,7 @@
struct rq_entry {
struct list_head list;
+ uint32_t recover_seq;
int nodeid;
struct dlm_message request;
};
@@ -41,6 +42,7 @@ void dlm_add_requestqueue(struct dlm_ls *ls, int nodeid, struct dlm_message *ms)
return;
}
+ e->recover_seq = ls->ls_recover_seq & 0xFFFFFFFF;
e->nodeid = nodeid;
memcpy(&e->request, ms, ms->m_header.h_length);
@@ -63,6 +65,7 @@ void dlm_add_requestqueue(struct dlm_ls *ls, int nodeid, struct dlm_message *ms)
int dlm_process_requestqueue(struct dlm_ls *ls)
{
struct rq_entry *e;
+ struct dlm_message *ms;
int error = 0;
mutex_lock(&ls->ls_requestqueue_mutex);
@@ -76,7 +79,15 @@ int dlm_process_requestqueue(struct dlm_ls *ls)
e = list_entry(ls->ls_requestqueue.next, struct rq_entry, list);
mutex_unlock(&ls->ls_requestqueue_mutex);
- dlm_receive_message_saved(ls, &e->request);
+ ms = &e->request;
+
+ log_limit(ls, "dlm_process_requestqueue msg %d from %d "
+ "lkid %x remid %x result %d seq %u",
+ ms->m_type, ms->m_header.h_nodeid,
+ ms->m_lkid, ms->m_remid, ms->m_result,
+ e->recover_seq);
+
+ dlm_receive_message_saved(ls, &e->request, e->recover_seq);
mutex_lock(&ls->ls_requestqueue_mutex);
list_del(&e->list);
@@ -138,35 +149,7 @@ static int purge_request(struct dlm_ls *ls, struct dlm_message *ms, int nodeid)
if (!dlm_no_directory(ls))
return 0;
- /* with no directory, the master is likely to change as a part of
- recovery; requests to/from the defunct master need to be purged */
-
- switch (type) {
- case DLM_MSG_REQUEST:
- case DLM_MSG_CONVERT:
- case DLM_MSG_UNLOCK:
- case DLM_MSG_CANCEL:
- /* we're no longer the master of this resource, the sender
- will resend to the new master (see waiter_needs_recovery) */
-
- if (dlm_hash2nodeid(ls, ms->m_hash) != dlm_our_nodeid())
- return 1;
- break;
-
- case DLM_MSG_REQUEST_REPLY:
- case DLM_MSG_CONVERT_REPLY:
- case DLM_MSG_UNLOCK_REPLY:
- case DLM_MSG_CANCEL_REPLY:
- case DLM_MSG_GRANT:
- /* this reply is from the former master of the resource,
- we'll resend to the new master if needed */
-
- if (dlm_hash2nodeid(ls, ms->m_hash) != nodeid)
- return 1;
- break;
- }
-
- return 0;
+ return 1;
}
void dlm_purge_requestqueue(struct dlm_ls *ls)
diff --git a/fs/dlm/user.c b/fs/dlm/user.c
index 66d6c16bf44..142e21655ee 100644
--- a/fs/dlm/user.c
+++ b/fs/dlm/user.c
@@ -24,6 +24,7 @@
#include "lock.h"
#include "lvb_table.h"
#include "user.h"
+#include "ast.h"
static const char name_prefix[] = "dlm";
static const struct file_operations device_fops;
@@ -152,19 +153,16 @@ static void compat_output(struct dlm_lock_result *res,
not related to the lifetime of the lkb struct which is managed
entirely by refcount. */
-static int lkb_is_endoflife(struct dlm_lkb *lkb, int sb_status, int type)
+static int lkb_is_endoflife(int mode, int status)
{
- switch (sb_status) {
+ switch (status) {
case -DLM_EUNLOCK:
return 1;
case -DLM_ECANCEL:
case -ETIMEDOUT:
case -EDEADLK:
- if (lkb->lkb_grmode == DLM_LOCK_IV)
- return 1;
- break;
case -EAGAIN:
- if (type == AST_COMP && lkb->lkb_grmode == DLM_LOCK_IV)
+ if (mode == DLM_LOCK_IV)
return 1;
break;
}
@@ -174,12 +172,13 @@ static int lkb_is_endoflife(struct dlm_lkb *lkb, int sb_status, int type)
/* we could possibly check if the cancel of an orphan has resulted in the lkb
being removed and then remove that lkb from the orphans list and free it */
-void dlm_user_add_ast(struct dlm_lkb *lkb, int type, int mode)
+void dlm_user_add_ast(struct dlm_lkb *lkb, uint32_t flags, int mode,
+ int status, uint32_t sbflags, uint64_t seq)
{
struct dlm_ls *ls;
struct dlm_user_args *ua;
struct dlm_user_proc *proc;
- int eol = 0, ast_type;
+ int rv;
if (lkb->lkb_flags & (DLM_IFL_ORPHAN | DLM_IFL_DEAD))
return;
@@ -200,49 +199,29 @@ void dlm_user_add_ast(struct dlm_lkb *lkb, int type, int mode)
ua = lkb->lkb_ua;
proc = ua->proc;
- if (type == AST_BAST && ua->bastaddr == NULL)
+ if ((flags & DLM_CB_BAST) && ua->bastaddr == NULL)
goto out;
+ if ((flags & DLM_CB_CAST) && lkb_is_endoflife(mode, status))
+ lkb->lkb_flags |= DLM_IFL_ENDOFLIFE;
+
spin_lock(&proc->asts_spin);
- ast_type = lkb->lkb_ast_type;
- lkb->lkb_ast_type |= type;
- if (type == AST_BAST)
- lkb->lkb_bastmode = mode;
- else
- lkb->lkb_castmode = mode;
+ rv = dlm_add_lkb_callback(lkb, flags, mode, status, sbflags, seq);
+ if (rv < 0) {
+ spin_unlock(&proc->asts_spin);
+ goto out;
+ }
- if (!ast_type) {
+ if (list_empty(&lkb->lkb_cb_list)) {
kref_get(&lkb->lkb_ref);
- list_add_tail(&lkb->lkb_astqueue, &proc->asts);
- lkb->lkb_ast_first = type;
+ list_add_tail(&lkb->lkb_cb_list, &proc->asts);
wake_up_interruptible(&proc->wait);
}
- if (type == AST_COMP && (ast_type & AST_COMP))
- log_debug(ls, "ast overlap %x status %x %x",
- lkb->lkb_id, ua->lksb.sb_status, lkb->lkb_flags);
-
- eol = lkb_is_endoflife(lkb, ua->lksb.sb_status, type);
- if (eol) {
- lkb->lkb_flags |= DLM_IFL_ENDOFLIFE;
- }
-
- /* We want to copy the lvb to userspace when the completion
- ast is read if the status is 0, the lock has an lvb and
- lvb_ops says we should. We could probably have set_lvb_lock()
- set update_user_lvb instead and not need old_mode */
-
- if ((lkb->lkb_ast_type & AST_COMP) &&
- (lkb->lkb_lksb->sb_status == 0) &&
- lkb->lkb_lksb->sb_lvbptr &&
- dlm_lvb_operations[ua->old_mode + 1][lkb->lkb_grmode + 1])
- ua->update_user_lvb = 1;
- else
- ua->update_user_lvb = 0;
-
spin_unlock(&proc->asts_spin);
- if (eol) {
+ if (lkb->lkb_flags & DLM_IFL_ENDOFLIFE) {
+ /* N.B. spin_lock locks_spin, not asts_spin */
spin_lock(&proc->locks_spin);
if (!list_empty(&lkb->lkb_ownqueue)) {
list_del_init(&lkb->lkb_ownqueue);
@@ -413,8 +392,9 @@ static int device_create_lockspace(struct dlm_lspace_params *params)
if (!capable(CAP_SYS_ADMIN))
return -EPERM;
- error = dlm_new_lockspace(params->name, strlen(params->name),
- &lockspace, params->flags, DLM_USER_LVB_LEN);
+ error = dlm_new_lockspace(params->name, NULL, params->flags,
+ DLM_USER_LVB_LEN, NULL, NULL, NULL,
+ &lockspace);
if (error)
return error;
@@ -513,7 +493,6 @@ static ssize_t device_write(struct file *file, const char __user *buf,
{
struct dlm_user_proc *proc = file->private_data;
struct dlm_write_request *kbuf;
- sigset_t tmpsig, allsigs;
int error;
#ifdef CONFIG_COMPAT
@@ -523,6 +502,13 @@ static ssize_t device_write(struct file *file, const char __user *buf,
#endif
return -EINVAL;
+ /*
+ * can't compare against COMPAT/dlm_write_request32 because
+ * we don't yet know if is64bit is zero
+ */
+ if (count > sizeof(struct dlm_write_request) + DLM_RESNAME_MAXLEN)
+ return -EINVAL;
+
kbuf = kzalloc(count + 1, GFP_NOFS);
if (!kbuf)
return -ENOMEM;
@@ -570,9 +556,6 @@ static ssize_t device_write(struct file *file, const char __user *buf,
goto out_free;
}
- sigfillset(&allsigs);
- sigprocmask(SIG_BLOCK, &allsigs, &tmpsig);
-
error = -EINVAL;
switch (kbuf->cmd)
@@ -580,7 +563,7 @@ static ssize_t device_write(struct file *file, const char __user *buf,
case DLM_USER_LOCK:
if (!proc) {
log_print("no locking on control device");
- goto out_sig;
+ goto out_free;
}
error = device_user_lock(proc, &kbuf->i.lock);
break;
@@ -588,7 +571,7 @@ static ssize_t device_write(struct file *file, const char __user *buf,
case DLM_USER_UNLOCK:
if (!proc) {
log_print("no locking on control device");
- goto out_sig;
+ goto out_free;
}
error = device_user_unlock(proc, &kbuf->i.lock);
break;
@@ -596,7 +579,7 @@ static ssize_t device_write(struct file *file, const char __user *buf,
case DLM_USER_DEADLOCK:
if (!proc) {
log_print("no locking on control device");
- goto out_sig;
+ goto out_free;
}
error = device_user_deadlock(proc, &kbuf->i.lock);
break;
@@ -604,7 +587,7 @@ static ssize_t device_write(struct file *file, const char __user *buf,
case DLM_USER_CREATE_LOCKSPACE:
if (proc) {
log_print("create/remove only on control device");
- goto out_sig;
+ goto out_free;
}
error = device_create_lockspace(&kbuf->i.lspace);
break;
@@ -612,7 +595,7 @@ static ssize_t device_write(struct file *file, const char __user *buf,
case DLM_USER_REMOVE_LOCKSPACE:
if (proc) {
log_print("create/remove only on control device");
- goto out_sig;
+ goto out_free;
}
error = device_remove_lockspace(&kbuf->i.lspace);
break;
@@ -620,7 +603,7 @@ static ssize_t device_write(struct file *file, const char __user *buf,
case DLM_USER_PURGE:
if (!proc) {
log_print("no locking on control device");
- goto out_sig;
+ goto out_free;
}
error = device_user_purge(proc, &kbuf->i.purge);
break;
@@ -630,9 +613,6 @@ static ssize_t device_write(struct file *file, const char __user *buf,
kbuf->cmd);
}
- out_sig:
- sigprocmask(SIG_SETMASK, &tmpsig, NULL);
- recalc_sigpending();
out_free:
kfree(kbuf);
return error;
@@ -673,15 +653,11 @@ static int device_close(struct inode *inode, struct file *file)
{
struct dlm_user_proc *proc = file->private_data;
struct dlm_ls *ls;
- sigset_t tmpsig, allsigs;
ls = dlm_find_lockspace_local(proc->lockspace);
if (!ls)
return -ENOENT;
- sigfillset(&allsigs);
- sigprocmask(SIG_BLOCK, &allsigs, &tmpsig);
-
set_bit(DLM_PROC_FLAGS_CLOSING, &proc->flags);
dlm_clear_proc_locks(ls, proc);
@@ -699,14 +675,12 @@ static int device_close(struct inode *inode, struct file *file)
/* FIXME: AUTOFREE: if this ls is no longer used do
device_remove_lockspace() */
- sigprocmask(SIG_SETMASK, &tmpsig, NULL);
- recalc_sigpending();
-
return 0;
}
-static int copy_result_to_user(struct dlm_user_args *ua, int compat, int type,
- int mode, char __user *buf, size_t count)
+static int copy_result_to_user(struct dlm_user_args *ua, int compat,
+ uint32_t flags, int mode, int copy_lvb,
+ char __user *buf, size_t count)
{
#ifdef CONFIG_COMPAT
struct dlm_lock_result32 result32;
@@ -730,7 +704,7 @@ static int copy_result_to_user(struct dlm_user_args *ua, int compat, int type,
notes that a new blocking AST address and parameter are set even if
the conversion fails, so maybe we should just do that. */
- if (type == AST_BAST) {
+ if (flags & DLM_CB_BAST) {
result.user_astaddr = ua->bastaddr;
result.user_astparam = ua->bastparam;
result.bast_mode = mode;
@@ -750,8 +724,7 @@ static int copy_result_to_user(struct dlm_user_args *ua, int compat, int type,
/* copy lvb to userspace if there is one, it's been updated, and
the user buffer has space for it */
- if (ua->update_user_lvb && ua->lksb.sb_lvbptr &&
- count >= len + DLM_USER_LVB_LEN) {
+ if (copy_lvb && ua->lksb.sb_lvbptr && count >= len + DLM_USER_LVB_LEN) {
if (copy_to_user(buf+len, ua->lksb.sb_lvbptr,
DLM_USER_LVB_LEN)) {
error = -EFAULT;
@@ -801,13 +774,12 @@ static ssize_t device_read(struct file *file, char __user *buf, size_t count,
struct dlm_user_proc *proc = file->private_data;
struct dlm_lkb *lkb;
DECLARE_WAITQUEUE(wait, current);
- int error = 0, removed;
- int ret_type, ret_mode;
- int bastmode, castmode, do_bast, do_cast;
+ struct dlm_callback cb;
+ int rv, resid, copy_lvb = 0;
if (count == sizeof(struct dlm_device_version)) {
- error = copy_version_to_user(buf, count);
- return error;
+ rv = copy_version_to_user(buf, count);
+ return rv;
}
if (!proc) {
@@ -854,92 +826,57 @@ static ssize_t device_read(struct file *file, char __user *buf, size_t count,
}
}
- /* there may be both completion and blocking asts to return for
- the lkb, don't remove lkb from asts list unless no asts remain */
-
- lkb = list_entry(proc->asts.next, struct dlm_lkb, lkb_astqueue);
-
- removed = 0;
- ret_type = 0;
- ret_mode = 0;
- do_bast = lkb->lkb_ast_type & AST_BAST;
- do_cast = lkb->lkb_ast_type & AST_COMP;
- bastmode = lkb->lkb_bastmode;
- castmode = lkb->lkb_castmode;
-
- /* when both are queued figure out which to do first and
- switch first so the other goes in the next read */
-
- if (do_cast && do_bast) {
- if (lkb->lkb_ast_first == AST_COMP) {
- ret_type = AST_COMP;
- ret_mode = castmode;
- lkb->lkb_ast_type &= ~AST_COMP;
- lkb->lkb_ast_first = AST_BAST;
- } else {
- ret_type = AST_BAST;
- ret_mode = bastmode;
- lkb->lkb_ast_type &= ~AST_BAST;
- lkb->lkb_ast_first = AST_COMP;
- }
- } else {
- ret_type = lkb->lkb_ast_first;
- ret_mode = (ret_type == AST_COMP) ? castmode : bastmode;
- lkb->lkb_ast_type &= ~ret_type;
- lkb->lkb_ast_first = 0;
- }
-
- /* if we're doing a bast but the bast is unnecessary, then
- switch to do nothing or do a cast if that was needed next */
+ /* if we empty lkb_callbacks, we don't want to unlock the spinlock
+ without removing lkb_cb_list; so empty lkb_cb_list is always
+ consistent with empty lkb_callbacks */
- if ((ret_type == AST_BAST) &&
- dlm_modes_compat(bastmode, lkb->lkb_castmode_done)) {
- ret_type = 0;
- ret_mode = 0;
+ lkb = list_entry(proc->asts.next, struct dlm_lkb, lkb_cb_list);
- if (do_cast) {
- ret_type = AST_COMP;
- ret_mode = castmode;
- lkb->lkb_ast_type &= ~AST_COMP;
- lkb->lkb_ast_first = 0;
- }
+ rv = dlm_rem_lkb_callback(lkb->lkb_resource->res_ls, lkb, &cb, &resid);
+ if (rv < 0) {
+ /* this shouldn't happen; lkb should have been removed from
+ list when resid was zero */
+ log_print("dlm_rem_lkb_callback empty %x", lkb->lkb_id);
+ list_del_init(&lkb->lkb_cb_list);
+ spin_unlock(&proc->asts_spin);
+ /* removes ref for proc->asts, may cause lkb to be freed */
+ dlm_put_lkb(lkb);
+ goto try_another;
}
+ if (!resid)
+ list_del_init(&lkb->lkb_cb_list);
+ spin_unlock(&proc->asts_spin);
- if (lkb->lkb_ast_first != lkb->lkb_ast_type) {
- log_print("device_read %x ast_first %x ast_type %x",
- lkb->lkb_id, lkb->lkb_ast_first, lkb->lkb_ast_type);
+ if (cb.flags & DLM_CB_SKIP) {
+ /* removes ref for proc->asts, may cause lkb to be freed */
+ if (!resid)
+ dlm_put_lkb(lkb);
+ goto try_another;
}
- if (!lkb->lkb_ast_type) {
- list_del(&lkb->lkb_astqueue);
- removed = 1;
- }
- spin_unlock(&proc->asts_spin);
+ if (cb.flags & DLM_CB_CAST) {
+ int old_mode, new_mode;
+
+ old_mode = lkb->lkb_last_cast.mode;
+ new_mode = cb.mode;
- if (ret_type) {
- error = copy_result_to_user(lkb->lkb_ua,
- test_bit(DLM_PROC_FLAGS_COMPAT, &proc->flags),
- ret_type, ret_mode, buf, count);
+ if (!cb.sb_status && lkb->lkb_lksb->sb_lvbptr &&
+ dlm_lvb_operations[old_mode + 1][new_mode + 1])
+ copy_lvb = 1;
- if (ret_type == AST_COMP)
- lkb->lkb_castmode_done = castmode;
- if (ret_type == AST_BAST)
- lkb->lkb_bastmode_done = bastmode;
+ lkb->lkb_lksb->sb_status = cb.sb_status;
+ lkb->lkb_lksb->sb_flags = cb.sb_flags;
}
- /* removes reference for the proc->asts lists added by
- dlm_user_add_ast() and may result in the lkb being freed */
+ rv = copy_result_to_user(lkb->lkb_ua,
+ test_bit(DLM_PROC_FLAGS_COMPAT, &proc->flags),
+ cb.flags, cb.mode, copy_lvb, buf, count);
- if (removed)
+ /* removes ref for proc->asts, may cause lkb to be freed */
+ if (!resid)
dlm_put_lkb(lkb);
- /* the bast that was queued was eliminated (see unnecessary above),
- leaving nothing to return */
-
- if (!ret_type)
- goto try_another;
-
- return error;
+ return rv;
}
static unsigned int device_poll(struct file *file, poll_table *wait)
diff --git a/fs/dlm/user.h b/fs/dlm/user.h
index f196091dd7f..00499ab8835 100644
--- a/fs/dlm/user.h
+++ b/fs/dlm/user.h
@@ -9,7 +9,8 @@
#ifndef __USER_DOT_H__
#define __USER_DOT_H__
-void dlm_user_add_ast(struct dlm_lkb *lkb, int type, int mode);
+void dlm_user_add_ast(struct dlm_lkb *lkb, uint32_t flags, int mode,
+ int status, uint32_t sbflags, uint64_t seq);
int dlm_user_init(void);
void dlm_user_exit(void);
int dlm_device_deregister(struct dlm_ls *ls);
generated by cgit v1.2.3-70-g09d2 (git 2.46.0) at 2025-11-04 20:18:28 +0000