summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorKrishnan Parthasarathi <kp@gluster.com>2012-02-21 16:28:13 +0530
committerAnand Avati <avati@redhat.com>2013-01-23 09:17:00 -0800
commit1ab086d8630687985fb412f4093a53d3e3e9aca1 (patch)
treeb69575ac19b190081cfaa92270fcb53ffab5a3c7
parent91ac9f97417790f439702c0297bca953ece597c8 (diff)
afr: Modified book-keeping structures for entrylks
* There are upto 3 entry lockees that may be needed to perform entrylk'ing in posix dir-write operations. * For eg, rmdir ("/a/b") needs to acquire locks on two entities, - entrylk ("/a", "b") - entrylk ("/a/b", null) * Changed existing entrylk/rename/selfheal (entrylk) transactions to use the new book-keeping structures * Fixed few issues in afr_trace_entry_lk{in,out} functions. Tracing is now aware of the new entry lockee structure. Implementation notes: * Changed 'cookie' sent in stack_wind to encode lockee_entity_no and subvol_no. cookie is a non-negative integer such that 0 <= cookie < replica_count, When more than one lock is being acquired across the subvolumes, cookie % replica_count gives the subvol_no cookie / replica_count gives the lockee_entity_no. Change-Id: Idbf41803387a7d59a0f7fcb1453d91cea74da153 BUG: 765564 Signed-off-by: Krishnan Parthasarathi <kp@gluster.com> Reviewed-on: http://review.gluster.org/2828 Tested-by: Gluster Build System <jenkins@build.gluster.com> Reviewed-by: Anand Avati <avati@redhat.com>
-rw-r--r--tests/bugs/bug-765564.t69
-rw-r--r--xlators/cluster/afr/src/afr-common.c8
-rw-r--r--xlators/cluster/afr/src/afr-dir-write.c209
-rw-r--r--xlators/cluster/afr/src/afr-lk-common.c680
-rw-r--r--xlators/cluster/afr/src/afr-self-heal-common.c44
-rw-r--r--xlators/cluster/afr/src/afr-transaction.c8
-rw-r--r--xlators/cluster/afr/src/afr.h23
7 files changed, 581 insertions, 460 deletions
diff --git a/tests/bugs/bug-765564.t b/tests/bugs/bug-765564.t
new file mode 100644
index 00000000000..cf00f520f14
--- /dev/null
+++ b/tests/bugs/bug-765564.t
@@ -0,0 +1,69 @@
+#!/bin/bash
+
+. $(dirname $0)/../include.rc
+
+TEST glusterd
+TEST pidof glusterd
+
+## Start and create a volume
+mkdir -p ${B0}/${V0}-0
+mkdir -p ${B0}/${V0}-1
+TEST $CLI volume create $V0 replica 2 $H0:$B0/${V0}-{0,1}
+
+TEST $CLI volume set $V0 performance.io-cache off;
+TEST $CLI volume set $V0 performance.write-behind off;
+TEST $CLI volume set $V0 performance.stat-prefetch off
+
+TEST $CLI volume start $V0;
+
+## Mount native
+TEST glusterfs --volfile-server=$H0 --volfile-id=$V0 $M0
+
+#returns success if 'olddir' is absent
+#'olddir' must be absent in both replicas
+function rm_succeeded () {
+ local dir1=$1
+ [[ -d $H0:$B0/${V0}-0/$dir1 || -d $H0:$B0/${V0}-1/$dir1 ]] && return 0
+ return 1
+}
+
+# returns successes if 'newdir' is present
+#'newdir' must be present in both replicas
+function mv_succeeded () {
+ local dir1=$1
+ [[ -d $H0:$B0/${V0}-0/$dir1 && -d $H0:$B0/${V0}-1/$dir1 ]] && return 1
+ return 0
+}
+
+# returns zero on success
+# Only one of rm and mv can succeed. This is captured by the XOR below
+
+function chk_backend_consistency(){
+ local dir1=$1
+ local dir2=$2
+ local rm_status=rm_succeeded $dir1
+ local mv_status=mv_succeeded $dir2
+ [[ ( $rm_status && ! $mv_status ) || ( ! $rm_status && $mv_status ) ]] && return 0
+ return 1
+}
+
+#concurrent removal/rename of dirs
+function rm_mv_correctness () {
+ ret=0
+ for i in {1..100}; do
+ mkdir $M0/"dir"$i
+ rmdir $M0/"dir"$i &
+ mv $M0/"dir"$i $M0/"adir"$i &
+ wait
+ tmp_ret=$(chk_backend_consistency "dir"$i "adir"$i)
+ (( ret += tmp_ret ))
+ rm -rf $M0/"dir"$i
+ rm -rf $M0/"adir"$i
+ done
+ return $ret
+}
+
+TEST rm_mv_correctness;
+TEST umount $M0;
+cleanup;
+
diff --git a/xlators/cluster/afr/src/afr-common.c b/xlators/cluster/afr/src/afr-common.c
index dbd64cb246b..5c4a649b95e 100644
--- a/xlators/cluster/afr/src/afr-common.c
+++ b/xlators/cluster/afr/src/afr-common.c
@@ -853,10 +853,9 @@ afr_local_transaction_cleanup (afr_local_t *local, xlator_t *this)
GF_FREE (local->internal_lock.inode_locked_nodes);
- GF_FREE (local->internal_lock.entry_locked_nodes);
-
GF_FREE (local->internal_lock.lower_locked_nodes);
+ afr_entry_lockee_cleanup (&local->internal_lock);
GF_FREE (local->transaction.pre_op);
GF_FREE (local->transaction.eager_lock);
@@ -3913,11 +3912,6 @@ afr_internal_lock_init (afr_internal_lock_t *lk, size_t child_count,
if (NULL == lk->inode_locked_nodes)
goto out;
- lk->entry_locked_nodes = GF_CALLOC (sizeof (*lk->entry_locked_nodes),
- child_count, gf_afr_mt_char);
- if (NULL == lk->entry_locked_nodes)
- goto out;
-
lk->locked_nodes = GF_CALLOC (sizeof (*lk->locked_nodes),
child_count, gf_afr_mt_char);
if (NULL == lk->locked_nodes)
diff --git a/xlators/cluster/afr/src/afr-dir-write.c b/xlators/cluster/afr/src/afr-dir-write.c
index d86b08b8784..a988ea2a829 100644
--- a/xlators/cluster/afr/src/afr-dir-write.c
+++ b/xlators/cluster/afr/src/afr-dir-write.c
@@ -418,11 +418,12 @@ afr_create (call_frame_t *frame, xlator_t *this,
loc_t *loc, int32_t flags, mode_t mode,
mode_t umask, fd_t *fd, dict_t *params)
{
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- call_frame_t *transaction_frame = NULL;
- int ret = -1;
- int op_errno = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = 0;
VALIDATE_OR_GOTO (frame, out);
VALIDATE_OR_GOTO (this, out);
@@ -473,7 +474,17 @@ afr_create (call_frame_t *frame, xlator_t *this,
local->transaction.main_frame = frame;
local->transaction.basename = AFR_BASENAME (loc->path);
+ int_lock = &local->internal_lock;
+ int_lock->lockee_count = 0;
+ ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
+ &local->transaction.parent_loc,
+ local->transaction.basename,
+ priv->child_count);
+ if (ret)
+ goto out;
+
+ int_lock->lockee_count++;
ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
if (ret < 0) {
op_errno = -ret;
@@ -612,11 +623,12 @@ int
afr_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
dev_t dev, mode_t umask, dict_t *params)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- call_frame_t * transaction_frame = NULL;
- int ret = -1;
- int op_errno = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = 0;
VALIDATE_OR_GOTO (frame, out);
VALIDATE_OR_GOTO (this, out);
@@ -666,7 +678,17 @@ afr_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
local->transaction.main_frame = frame;
local->transaction.basename = AFR_BASENAME (loc->path);
+ int_lock = &local->internal_lock;
+
+ int_lock->lockee_count = 0;
+ ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
+ &local->transaction.parent_loc,
+ local->transaction.basename,
+ priv->child_count);
+ if (ret)
+ goto out;
+ int_lock->lockee_count++;
ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
if (ret < 0) {
op_errno = -ret;
@@ -802,16 +824,16 @@ afr_mkdir_done (call_frame_t *frame, xlator_t *this)
return 0;
}
-
int
afr_mkdir (call_frame_t *frame, xlator_t *this,
loc_t *loc, mode_t mode, mode_t umask, dict_t *params)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- call_frame_t * transaction_frame = NULL;
- int ret = -1;
- int op_errno = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = 0;
VALIDATE_OR_GOTO (frame, out);
VALIDATE_OR_GOTO (this, out);
@@ -859,7 +881,17 @@ afr_mkdir (call_frame_t *frame, xlator_t *this,
local->transaction.main_frame = frame;
local->transaction.basename = AFR_BASENAME (loc->path);
+ int_lock = &local->internal_lock;
+
+ int_lock->lockee_count = 0;
+ ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
+ &local->transaction.parent_loc,
+ local->transaction.basename,
+ priv->child_count);
+ if (ret)
+ goto out;
+ int_lock->lockee_count++;
ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
if (ret < 0) {
op_errno = -ret;
@@ -966,7 +998,8 @@ afr_link_wind (call_frame_t *frame, xlator_t *this)
for (i = 0; i < priv->child_count; i++) {
if (local->transaction.pre_op[i]) {
- STACK_WIND_COOKIE (frame, afr_link_wind_cbk, (void *) (long) i,
+ STACK_WIND_COOKIE (frame, afr_link_wind_cbk,
+ (void *) (long) i,
priv->children[i],
priv->children[i]->fops->link,
&local->loc,
@@ -998,11 +1031,12 @@ int
afr_link (call_frame_t *frame, xlator_t *this,
loc_t *oldloc, loc_t *newloc, dict_t *xdata)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- call_frame_t * transaction_frame = NULL;
- int ret = -1;
- int op_errno = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = 0;
VALIDATE_OR_GOTO (frame, out);
VALIDATE_OR_GOTO (this, out);
@@ -1048,13 +1082,22 @@ afr_link (call_frame_t *frame, xlator_t *this,
local->transaction.main_frame = frame;
local->transaction.basename = AFR_BASENAME (newloc->path);
+ int_lock = &local->internal_lock;
+ int_lock->lockee_count = 0;
+ ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
+ &local->transaction.parent_loc,
+ local->transaction.basename,
+ priv->child_count);
+ if (ret)
+ goto out;
+
+ int_lock->lockee_count++;
ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
if (ret < 0) {
op_errno = -ret;
goto out;
}
-
ret = 0;
out:
if (ret < 0) {
@@ -1190,11 +1233,12 @@ int
afr_symlink (call_frame_t *frame, xlator_t *this,
const char *linkpath, loc_t *loc, mode_t umask, dict_t *params)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- call_frame_t * transaction_frame = NULL;
- int ret = -1;
- int op_errno = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = 0;
VALIDATE_OR_GOTO (frame, out);
VALIDATE_OR_GOTO (this, out);
@@ -1242,7 +1286,17 @@ afr_symlink (call_frame_t *frame, xlator_t *this,
local->transaction.main_frame = frame;
local->transaction.basename = AFR_BASENAME (loc->path);
+ int_lock = &local->internal_lock;
+ int_lock->lockee_count = 0;
+ ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
+ &local->transaction.parent_loc,
+ local->transaction.basename,
+ priv->child_count);
+ if (ret)
+ goto out;
+
+ int_lock->lockee_count++;
ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
if (ret < 0) {
op_errno = -ret;
@@ -1391,11 +1445,13 @@ int
afr_rename (call_frame_t *frame, xlator_t *this,
loc_t *oldloc, loc_t *newloc, dict_t *xdata)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- call_frame_t * transaction_frame = NULL;
- int ret = -1;
- int op_errno = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = 0;
+ int nlockee = 0;
VALIDATE_OR_GOTO (frame, out);
VALIDATE_OR_GOTO (this, out);
@@ -1439,6 +1495,38 @@ afr_rename (call_frame_t *frame, xlator_t *this,
local->transaction.main_frame = frame;
local->transaction.basename = AFR_BASENAME (oldloc->path);
local->transaction.new_basename = AFR_BASENAME (newloc->path);
+ int_lock = &local->internal_lock;
+
+ int_lock->lockee_count = nlockee = 0;
+ ret = afr_init_entry_lockee (&int_lock->lockee[nlockee], local,
+ &local->transaction.new_parent_loc,
+ local->transaction.new_basename,
+ priv->child_count);
+ if (ret)
+ goto out;
+
+ nlockee++;
+ ret = afr_init_entry_lockee (&int_lock->lockee[nlockee], local,
+ &local->transaction.parent_loc,
+ local->transaction.basename,
+ priv->child_count);
+ if (ret)
+ goto out;
+
+ nlockee++;
+ if (local->newloc.inode && IA_ISDIR (local->newloc.inode->ia_type)) {
+ ret = afr_init_entry_lockee (&int_lock->lockee[nlockee], local,
+ &local->newloc,
+ NULL,
+ priv->child_count);
+ if (ret)
+ goto out;
+
+ nlockee++;
+ }
+ qsort (int_lock->lockee, nlockee, sizeof (*int_lock->lockee),
+ afr_entry_lockee_cmp);
+ int_lock->lockee_count = nlockee;
ret = afr_transaction (transaction_frame, this, AFR_ENTRY_RENAME_TRANSACTION);
if (ret < 0) {
@@ -1578,11 +1666,12 @@ int32_t
afr_unlink (call_frame_t *frame, xlator_t *this,
loc_t *loc, int xflag, dict_t *xdata)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- call_frame_t * transaction_frame = NULL;
- int ret = -1;
- int op_errno = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = 0;
VALIDATE_OR_GOTO (frame, out);
VALIDATE_OR_GOTO (this, out);
@@ -1621,7 +1710,17 @@ afr_unlink (call_frame_t *frame, xlator_t *this,
local->transaction.main_frame = frame;
local->transaction.basename = AFR_BASENAME (loc->path);
+ int_lock = &local->internal_lock;
+ int_lock->lockee_count = 0;
+ ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
+ &local->transaction.parent_loc,
+ local->transaction.basename,
+ priv->child_count);
+ if (ret)
+ goto out;
+
+ int_lock->lockee_count++;
ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
if (ret < 0) {
op_errno = -ret;
@@ -1768,11 +1867,13 @@ int
afr_rmdir (call_frame_t *frame, xlator_t *this,
loc_t *loc, int flags, dict_t *xdata)
{
- afr_private_t * priv = NULL;
- afr_local_t * local = NULL;
- call_frame_t * transaction_frame = NULL;
- int ret = -1;
- int op_errno = 0;
+ afr_private_t *priv = NULL;
+ afr_local_t *local = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ call_frame_t *transaction_frame = NULL;
+ int ret = -1;
+ int op_errno = 0;
+ int nlockee = 0;
VALIDATE_OR_GOTO (frame, out);
VALIDATE_OR_GOTO (this, out);
@@ -1809,6 +1910,28 @@ afr_rmdir (call_frame_t *frame, xlator_t *this,
local->transaction.main_frame = frame;
local->transaction.basename = AFR_BASENAME (loc->path);
+ int_lock = &local->internal_lock;
+
+ int_lock->lockee_count = nlockee = 0;
+ ret = afr_init_entry_lockee (&int_lock->lockee[nlockee], local,
+ &local->transaction.parent_loc,
+ local->transaction.basename,
+ priv->child_count);
+ if (ret)
+ goto out;
+
+ nlockee++;
+ ret = afr_init_entry_lockee (&int_lock->lockee[nlockee], local,
+ &local->loc,
+ NULL,
+ priv->child_count);
+ if (ret)
+ goto out;
+
+ nlockee++;
+ qsort (int_lock->lockee, nlockee, sizeof (*int_lock->lockee),
+ afr_entry_lockee_cmp);
+ int_lock->lockee_count = nlockee;
ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
if (ret < 0) {
diff --git a/xlators/cluster/afr/src/afr-lk-common.c b/xlators/cluster/afr/src/afr-lk-common.c
index 5e61be4d4df..ef183d985cb 100644
--- a/xlators/cluster/afr/src/afr-lk-common.c
+++ b/xlators/cluster/afr/src/afr-lk-common.c
@@ -55,7 +55,30 @@
} while (0);
int
-afr_lock_blocking (call_frame_t *frame, xlator_t *this, int child_index);
+afr_entry_lockee_cmp (const void *l1, const void *l2)
+{
+ const afr_entry_lockee_t *r1 = l1;
+ const afr_entry_lockee_t *r2 = l2;
+ int ret = 0;
+
+ ret = uuid_compare (r1->loc.gfid, r2->loc.gfid);
+ /*Entrylks with NULL basename are the 'smallest'*/
+ if (ret == 0) {
+ if (!r1->basename)
+ return -1;
+ if (!r2->basename)
+ return 1;
+ ret = strcmp (r1->basename, r2->basename);
+ }
+
+ if (ret <= 0)
+ return -1;
+ else
+ return 1;
+
+}
+
+int afr_lock_blocking (call_frame_t *frame, xlator_t *this, int child_index);
static int
afr_copy_locked_nodes (call_frame_t *frame, xlator_t *this);
@@ -332,10 +355,13 @@ static void
afr_trace_entrylk_in (call_frame_t *frame, xlator_t *this,
afr_lock_call_type_t lock_call_type,
afr_lock_op_type_t lk_op_type, const char *basename,
- int32_t child_index)
+ int32_t cookie)
{
afr_local_t *local = NULL;
afr_internal_lock_t *int_lock = NULL;
+ afr_private_t *priv = NULL;
+ int child_index = 0;
+ int lockee_no = 0;
char lock[256];
char lockee[256];
@@ -343,28 +369,40 @@ afr_trace_entrylk_in (call_frame_t *frame, xlator_t *this,
local = frame->local;
int_lock = &local->internal_lock;
+ priv = this->private;
+
+ if (!priv->entrylk_trace) {
+ return;
+ }
+ lockee_no = cookie / priv->child_count;
+ child_index = cookie % priv->child_count;
afr_print_entrylk (lock, 256, basename, &frame->root->lk_owner);
- afr_print_lockee (lockee, 256, &local->loc, local->fd, child_index);
+ afr_print_lockee (lockee, 256, &int_lock->lockee[lockee_no].loc, local->fd,
+ child_index);
afr_set_lock_call_type (lock_call_type, lock_call_type_str, int_lock);
gf_log (this->name, GF_LOG_INFO,
- "[%s %s] Lock={%s} Lockee={%s} Number={%llu}",
+ "[%s %s] Lock={%s} Lockee={%s} Number={%llu}, Cookie={%d}",
lock_call_type_str,
lk_op_type == AFR_LOCK_OP ? "LOCK REQUEST" : "UNLOCK REQUEST",
lock, lockee,
- (unsigned long long) int_lock->lock_number);
+ (unsigned long long) int_lock->lock_number,
+ cookie);
}
static void
afr_trace_entrylk_out (call_frame_t *frame, xlator_t *this,
afr_lock_call_type_t lock_call_type,
afr_lock_op_type_t lk_op_type, const char *basename,
- int op_ret, int op_errno, int32_t child_index)
+ int op_ret, int op_errno, int32_t cookie)
{
afr_internal_lock_t *int_lock = NULL;
afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int lockee_no = 0;
+ int child_index = 0;
char lock[256];
char lockee[256];
@@ -373,20 +411,30 @@ afr_trace_entrylk_out (call_frame_t *frame, xlator_t *this,
local = frame->local;
int_lock = &local->internal_lock;
+ priv = this->private;
- afr_print_lockee (lockee, 256, &local->loc, local->fd, child_index);
+ if (!priv->entrylk_trace) {
+ return;
+ }
+ lockee_no = cookie / priv->child_count;
+ child_index = cookie % priv->child_count;
+
+ afr_print_entrylk (lock, 256, basename, &frame->root->lk_owner);
+ afr_print_lockee (lockee, 256, &int_lock->lockee[lockee_no].loc, local->fd,
+ child_index);
afr_set_lock_call_type (lock_call_type, lock_call_type_str, int_lock);
afr_print_verdict (op_ret, op_errno, verdict);
gf_log (this->name, GF_LOG_INFO,
- "[%s %s] [%s] Lock={%s} Lockee={%s} Number={%llu}",
+ "[%s %s] [%s] Lock={%s} Lockee={%s} Number={%llu} Cookie={%d}",
lock_call_type_str,
lk_op_type == AFR_LOCK_OP ? "LOCK REPLY" : "UNLOCK REPLY",
verdict,
lock, lockee,
- (unsigned long long) int_lock->lock_number);
+ (unsigned long long) int_lock->lock_number,
+ cookie);
}
@@ -439,6 +487,47 @@ is_afr_lock_transaction (afr_local_t *local)
return ret;
}
+int
+afr_init_entry_lockee (afr_entry_lockee_t *lockee, afr_local_t *local,
+ loc_t *loc, char *basename, int child_count)
+{
+ int ret = -1;
+
+ loc_copy (&lockee->loc, loc);
+ lockee->basename = (basename)? gf_strdup (basename): NULL;
+ if (basename && !lockee->basename)
+ goto out;
+
+ lockee->locked_count = 0;
+ lockee->locked_nodes = GF_CALLOC (child_count,
+ sizeof (*lockee->locked_nodes),
+ gf_afr_mt_afr_node_character);
+
+ if (!lockee->locked_nodes)
+ goto out;
+
+ ret = 0;
+out:
+ return ret;
+
+}
+
+void
+afr_entry_lockee_cleanup (afr_internal_lock_t *int_lock)
+{
+ int i = 0;
+
+ for (i = 0; i < int_lock->lockee_count; i++) {
+ loc_wipe (&int_lock->lockee[i].loc);
+ if (int_lock->lockee[i].basename)
+ GF_FREE (int_lock->lockee[i].basename);
+ if (int_lock->lockee[i].locked_nodes)
+ GF_FREE (int_lock->lockee[i].locked_nodes);
+ }
+
+ return;
+}
+
static int
initialize_entrylk_variables (call_frame_t *frame, xlator_t *this)
{
@@ -456,8 +545,13 @@ initialize_entrylk_variables (call_frame_t *frame, xlator_t *this)
int_lock->lock_op_ret = -1;
int_lock->lock_op_errno = 0;
- for (i = 0; i < priv->child_count; i++) {
- int_lock->entry_locked_nodes[i] = 0;
+ for (i = 0; i < AFR_LOCKEE_COUNT_MAX; i++) {
+ if (!int_lock->lockee[i].locked_nodes)
+ break;
+ int_lock->lockee[i].locked_count = 0;
+ memset (int_lock->lockee[i].locked_nodes, 0,
+ sizeof (*int_lock->lockee[i].locked_nodes) *
+ priv->child_count);
}
return 0;
@@ -503,6 +597,18 @@ lower_path (loc_t *l1, const char *b1, loc_t *l2, const char *b2)
}
int
+afr_lockee_locked_nodes_count (afr_internal_lock_t *int_lock)
+{
+ int call_count = 0;
+ int i = 0;
+
+ for (i = 0; i < int_lock->lockee_count; i++)
+ call_count += int_lock->lockee[i].locked_count;
+
+ return call_count;
+}
+
+int
afr_locked_nodes_count (unsigned char *locked_nodes, int child_count)
{
@@ -689,13 +795,22 @@ afr_unlock_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
afr_local_t *local = NULL;
- int32_t child_index = (long)cookie;
+ afr_private_t *priv = NULL;
+ afr_internal_lock_t *int_lock = NULL;
+ int32_t child_index = 0;
+ int lockee_no = 0;
+
+ priv = this->private;
+ lockee_no = (int)((long) cookie) / priv->child_count;
+ child_index = (int) ((long) cookie) % priv->child_count;
local = frame->local;
+ int_lock = &local->internal_lock;
AFR_TRACE_ENTRYLK_OUT (frame, this, AFR_ENTRYLK_TRANSACTION,
- AFR_UNLOCK_OP, NULL, op_ret,
- op_errno, child_index);
+ AFR_UNLOCK_OP,
+ int_lock->lockee[lockee_no].basename, op_ret,
+ op_errno, (int) ((long)cookie));
if (op_ret < 0) {
gf_log (this->name, GF_LOG_ERROR,
@@ -703,6 +818,7 @@ afr_unlock_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
local->loc.path, child_index, strerror (op_errno));
}
+ int_lock->lockee[lockee_no].locked_nodes[child_index] &= LOCKED_NO;
afr_unlock_common_cbk (frame, cookie, this, op_ret, op_errno, NULL);
return 0;
@@ -711,24 +827,22 @@ afr_unlock_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
static int
afr_unlock_entrylk (call_frame_t *frame, xlator_t *this)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- const char *basename = NULL;
- loc_t *loc = NULL;
- int call_count = 0;
- int i = -1;
+ afr_internal_lock_t *int_lock = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int call_count = 0;
+ int index = 0;
+ int lockee_no = 0;
+ int copies = 0;
+ int i = -1;
local = frame->local;
int_lock = &local->internal_lock;
priv = this->private;
+ copies = priv->child_count;
- basename = int_lock->lk_basename;
- if (int_lock->lk_loc)
- loc = int_lock->lk_loc;
+ call_count = afr_lockee_locked_nodes_count (int_lock);
- call_count = afr_locked_nodes_count (int_lock->entry_locked_nodes,
- priv->child_count);
int_lock->lk_call_count = call_count;
if (!call_count){
@@ -738,18 +852,22 @@ afr_unlock_entrylk (call_frame_t *frame, xlator_t *this)
goto out;
}
- for (i = 0; i < priv->child_count; i++) {
- if (int_lock->entry_locked_nodes[i] & LOCKED_YES) {
- AFR_TRACE_ENTRYLK_IN (frame, this,
- AFR_ENTRYLK_NB_TRANSACTION,
- AFR_UNLOCK_OP, basename, i);
+ for (i = 0; i < int_lock->lockee_count * priv->child_count; i++) {
+ lockee_no = i / copies;
+ index = i % copies;
+ if (int_lock->lockee[lockee_no].locked_nodes[index] & LOCKED_YES) {
+ AFR_TRACE_ENTRYLK_IN (frame, this, AFR_ENTRYLK_NB_TRANSACTION,
+ AFR_UNLOCK_OP,
+ int_lock->lockee[lockee_no].basename,
+ i);
STACK_WIND_COOKIE (frame, afr_unlock_entrylk_cbk,
(void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->entrylk,
+ priv->children[index],
+ priv->children[index]->fops->entrylk,
this->name,
- loc, basename,
+ &int_lock->lockee[lockee_no].loc,
+ int_lock->lockee[lockee_no].basename,
ENTRYLK_UNLOCK, ENTRYLK_WRLCK, NULL);
if (!--call_count)
@@ -766,13 +884,20 @@ static int32_t
afr_lock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- int child_index = (long) cookie;
+ afr_internal_lock_t *int_lock = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ int cky = (long) cookie;
+ int child_index = 0;
+ int lockee_no = 0;
+ priv = this->private;
local = frame->local;
int_lock = &local->internal_lock;
+ child_index = ((int)cky) % priv->child_count;
+ lockee_no = ((int)cky) / priv->child_count;
+
LOCK (&frame->lock);
{
if (op_ret == -1) {
@@ -796,10 +921,17 @@ afr_lock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
afr_unlock (frame, this);
} else {
if (op_ret == 0) {
- int_lock->locked_nodes[child_index] |= LOCKED_YES;
- int_lock->lock_count++;
+ if (local->transaction.type == AFR_ENTRY_TRANSACTION ||
+ local->transaction.type == AFR_ENTRY_RENAME_TRANSACTION) {
+ int_lock->lockee[lockee_no].locked_nodes[child_index] |= LOCKED_YES;
+ int_lock->lockee[lockee_no].locked_count++;
+ int_lock->entrylk_lock_count++;
+ } else {
+ int_lock->locked_nodes[child_index] |= LOCKED_YES;
+ int_lock->lock_count++;
+ }
}
- afr_lock_blocking (frame, this, child_index + 1);
+ afr_lock_blocking (frame, this, cky + 1);
}
return 0;
@@ -819,79 +951,6 @@ afr_blocking_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
}
static int32_t
-afr_lock_lower_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
- int32_t op_ret, int32_t op_errno, dict_t *xdata)
-{
- afr_internal_lock_t *int_lock = NULL;
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- loc_t *lower = NULL;
- loc_t *higher = NULL;
- const char *higher_name = NULL;
- int child_index = (long) cookie;
-
- priv = this->private;
- local = frame->local;
- int_lock = &local->internal_lock;
-
- LOCK (&frame->lock);
- {
- if (op_ret == -1) {
- if (op_errno == ENOSYS) {
- /* return ENOTSUP */
-
- gf_log (this->name, GF_LOG_ERROR,
- "subvolume does not support locking. "
- "please load features/locks xlator on server");
-
- local->op_ret = op_ret;
- }
-
- local->op_errno = op_errno;
- }
- }
- UNLOCK (&frame->lock);
-
- if (op_ret != 0) {
- afr_copy_locked_nodes (frame, this);
- afr_unlock (frame, this);
- goto out;
- } else {
- int_lock->lower_locked_nodes[child_index] |= LOCKED_LOWER;
- int_lock->lock_count++;
- }
-
- /* The lower path has been locked. Now lock the higher path */
-
- lower = lower_path (&local->transaction.parent_loc,
- local->transaction.basename,
- &local->transaction.new_parent_loc,
- local->transaction.new_basename);
-
- higher = (lower == &local->transaction.parent_loc ?
- &local->transaction.new_parent_loc :
- &local->transaction.parent_loc);
-
- higher_name = (higher == &local->transaction.parent_loc ?
- local->transaction.basename :
- local->transaction.new_basename);
-
- AFR_TRACE_ENTRYLK_IN (frame, this, AFR_ENTRYLK_TRANSACTION,
- AFR_LOCK_OP, higher_name, child_index);
-
-
- STACK_WIND_COOKIE (frame, afr_lock_cbk,
- (void *) (long) child_index,
- priv->children[child_index],
- priv->children[child_index]->fops->entrylk,
- this->name, higher, higher_name,
- ENTRYLK_LOCK, ENTRYLK_WRLCK, NULL);
-
-out:
- return 0;
-}
-
-static int32_t
afr_blocking_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
int32_t op_ret, int32_t op_errno, dict_t *xdata)
{
@@ -925,10 +984,8 @@ afr_copy_locked_nodes (call_frame_t *frame, xlator_t *this)
case AFR_ENTRY_RENAME_TRANSACTION:
case AFR_ENTRY_TRANSACTION:
- memcpy (int_lock->entry_locked_nodes,
- int_lock->locked_nodes,
- priv->child_count);
- int_lock->entrylk_lock_count = int_lock->lock_count;
+ /*entrylk_count is being used in both non-blocking and blocking
+ * modes */
break;
}
@@ -936,21 +993,47 @@ afr_copy_locked_nodes (call_frame_t *frame, xlator_t *this)
}
+static inline gf_boolean_t
+afr_is_entrylk (afr_internal_lock_t *int_lock,
+ afr_transaction_type trans_type)
+{
+ gf_boolean_t is_entrylk = _gf_false;
+
+ if ((int_lock->transaction_lk_type == AFR_SELFHEAL_LK) &&
+ int_lock->selfheal_lk_type == AFR_ENTRY_SELF_HEAL_LK) {
+
+ is_entrylk = _gf_true;
+
+ } else if ((int_lock->transaction_lk_type == AFR_TRANSACTION_LK) &&
+ (trans_type == AFR_ENTRY_TRANSACTION ||
+ trans_type == AFR_ENTRY_RENAME_TRANSACTION)) {
+
+ is_entrylk = _gf_true;
+
+ } else {
+ is_entrylk = _gf_false;
+ }
+
+ return is_entrylk;
+}
+
int
-afr_lock_blocking (call_frame_t *frame, xlator_t *this, int child_index)
+afr_lock_blocking (call_frame_t *frame, xlator_t *this, int cookie)
{
afr_internal_lock_t *int_lock = NULL;
afr_local_t *local = NULL;
afr_private_t *priv = NULL;
- loc_t *lower = NULL;
- const char *lower_name = NULL;
struct gf_flock flock = {0,};
uint64_t ctx = 0;
int ret = 0;
+ int child_index = 0;
+ int lockee_no = 0;
- local = frame->local;
- int_lock = &local->internal_lock;
- priv = this->private;
+ local = frame->local;
+ int_lock = &local->internal_lock;
+ priv = this->private;
+ child_index = cookie % priv->child_count;
+ lockee_no = cookie / priv->child_count;
flock.l_start = int_lock->lk_flock.l_start;
flock.l_len = int_lock->lk_flock.l_len;
@@ -990,7 +1073,9 @@ afr_lock_blocking (call_frame_t *frame, xlator_t *this, int child_index)
}
if ((child_index == priv->child_count) &&
- int_lock->lock_count == 0) {
+ ((afr_is_entrylk (int_lock, local->transaction.type) &&
+ int_lock->entrylk_lock_count == 0) ||
+ (int_lock->lock_count == 0))){
gf_log (this->name, GF_LOG_INFO,
"unable to lock on even one child");
@@ -1006,8 +1091,9 @@ afr_lock_blocking (call_frame_t *frame, xlator_t *this, int child_index)
}
- if ((child_index == priv->child_count)
- || (int_lock->lock_count == int_lock->lk_expected_count)) {
+ if ((child_index == priv->child_count) ||
+ (int_lock->entrylk_lock_count == int_lock->lk_expected_count) ||
+ (int_lock->lock_count == int_lock->lk_expected_count)) {
/* we're done locking */
@@ -1055,43 +1141,23 @@ afr_lock_blocking (call_frame_t *frame, xlator_t *this, int child_index)
break;
case AFR_ENTRY_RENAME_TRANSACTION:
- {
- lower = lower_path (&local->transaction.parent_loc,
- local->transaction.basename,
- &local->transaction.new_parent_loc,
- local->transaction.new_basename);
-
- lower_name = (lower == &local->transaction.parent_loc ?
- local->transaction.basename :
- local->transaction.new_basename);
-
- AFR_TRACE_ENTRYLK_IN (frame, this, AFR_ENTRYLK_TRANSACTION,
- AFR_LOCK_OP, lower_name, child_index);
-
-
- STACK_WIND_COOKIE (frame, afr_lock_lower_cbk,
- (void *) (long) child_index,
- priv->children[child_index],
- priv->children[child_index]->fops->entrylk,
- this->name, lower, lower_name,
- ENTRYLK_LOCK, ENTRYLK_WRLCK, NULL);
-
- break;
- }
-
case AFR_ENTRY_TRANSACTION:
+ /*Accounting for child_index increments on 'down'
+ *and 'fd-less' children */
+ cookie = lockee_no * priv->child_count + child_index;
+
if (local->fd) {
- AFR_TRACE_ENTRYLK_IN (frame, this,
- AFR_ENTRYLK_TRANSACTION,
- AFR_LOCK_OP, local->transaction.basename,
- child_index);
+ AFR_TRACE_ENTRYLK_IN (frame, this, AFR_ENTRYLK_TRANSACTION,
+ AFR_LOCK_OP,
+ int_lock->lockee[lockee_no].basename,
+ cookie);
STACK_WIND_COOKIE (frame, afr_blocking_entrylk_cbk,
- (void *) (long) child_index,
+ (void *) (long) cookie,
priv->children[child_index],
priv->children[child_index]->fops->fentrylk,
this->name, local->fd,
- local->transaction.basename,
+ int_lock->lockee[lockee_no].basename,
ENTRYLK_LOCK, ENTRYLK_WRLCK, NULL);
} else {
AFR_TRACE_ENTRYLK_IN (frame, this,
@@ -1100,12 +1166,12 @@ afr_lock_blocking (call_frame_t *frame, xlator_t *this, int child_index)
child_index);
STACK_WIND_COOKIE (frame, afr_blocking_entrylk_cbk,
- (void *) (long) child_index,
+ (void *) (long) cookie,
priv->children[child_index],
priv->children[child_index]->fops->entrylk,
this->name,
- &local->transaction.parent_loc,
- local->transaction.basename,
+ &int_lock->lockee[lockee_no].loc,
+ int_lock->lockee[lockee_no].basename,
ENTRYLK_LOCK, ENTRYLK_WRLCK, NULL);
}
@@ -1134,11 +1200,12 @@ afr_blocking_lock (call_frame_t *frame, xlator_t *this)
break;
case AFR_ENTRY_RENAME_TRANSACTION:
+ case AFR_ENTRY_TRANSACTION:
up_count = afr_up_children_count (local->child_up,
priv->child_count);
- int_lock->lk_expected_count = 2 * up_count;
- //fallthrough
- case AFR_ENTRY_TRANSACTION:
+ int_lock->lk_call_count = int_lock->lk_expected_count
+ = (int_lock->lockee_count *
+ up_count);
initialize_entrylk_variables (frame, this);
break;
}
@@ -1154,14 +1221,25 @@ afr_nonblocking_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
{
afr_internal_lock_t *int_lock = NULL;
afr_local_t *local = NULL;
- int call_count = 0;
- int child_index = (long) cookie;
+ int call_count = 0;
+ int child_index = (long) cookie;
+ int copies = 0;
+ int index = 0;
+ int lockee_no = 0;
+ afr_private_t *priv = NULL;
+
+ priv = this->private;
+
+ copies = priv->child_count;
+ index = child_index % copies;
+ lockee_no = child_index / copies;
local = frame->local;
int_lock = &local->internal_lock;
AFR_TRACE_ENTRYLK_OUT (frame, this, AFR_ENTRYLK_TRANSACTION,
- AFR_LOCK_OP, NULL, op_ret,
+ AFR_LOCK_OP,
+ int_lock->lockee[lockee_no].basename, op_ret,
op_errno, (long) cookie);
if (op_ret < 0 ) {
@@ -1177,7 +1255,8 @@ afr_nonblocking_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
local->op_errno = op_errno;
}
} else if (op_ret == 0) {
- int_lock->entry_locked_nodes[child_index] |= LOCKED_YES;
+ int_lock->lockee[lockee_no].locked_nodes[index] |= LOCKED_YES;
+ int_lock->lockee[lockee_no].locked_count++;
int_lock->entrylk_lock_count++;
}
@@ -1229,25 +1308,23 @@ afr_mark_fd_open_on (afr_local_t *local, afr_fd_ctx_t *fd_ctx,
int
afr_nonblocking_entrylk (call_frame_t *frame, xlator_t *this)
{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- afr_fd_ctx_t *fd_ctx = NULL;
- const char *basename = NULL;
- loc_t *loc = NULL;
- int32_t call_count = 0;
+ afr_internal_lock_t *int_lock = NULL;
+ afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ afr_fd_ctx_t *fd_ctx = NULL;
+ int copies = 0;
+ int index = 0;
+ int lockee_no = 0;
+ int32_t call_count = 0;
int i = 0;
local = frame->local;
int_lock = &local->internal_lock;
priv = this->private;
+ copies = priv->child_count;
initialize_entrylk_variables (frame, this);
- basename = int_lock->lk_basename;
- if (int_lock->lk_loc)
- loc = int_lock->lk_loc;
-
if (local->fd) {
fd_ctx = afr_fd_ctx_get (local->fd, this);
if (!fd_ctx) {
@@ -1264,7 +1341,7 @@ afr_nonblocking_entrylk (call_frame_t *frame, xlator_t *this)
}
afr_mark_fd_open_on (local, fd_ctx, priv->child_count);
- call_count = internal_lock_count (frame, this);
+ call_count = int_lock->lockee_count * internal_lock_count (frame, this);
int_lock->lk_call_count = call_count;
int_lock->lk_expected_count = call_count;
@@ -1277,46 +1354,52 @@ afr_nonblocking_entrylk (call_frame_t *frame, xlator_t *this)
/* Send non-blocking entrylk calls only on up children
and where the fd has been opened */
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i] && local->fd_open_on[i]) {
- AFR_TRACE_ENTRYLK_IN (frame, this,
- AFR_ENTRYLK_NB_TRANSACTION,
- AFR_LOCK_OP, basename, i);
+ for (i = 0; i < int_lock->lockee_count*priv->child_count; i++) {
+ index = i%copies;
+ lockee_no = i/copies;
+ if (local->child_up[index] && local->fd_open_on[index]) {
+ AFR_TRACE_ENTRYLK_IN (frame, this, AFR_ENTRYLK_NB_TRANSACTION,
+ AFR_LOCK_OP,
+ int_lock->lockee[lockee_no].basename,
+ i);
STACK_WIND_COOKIE (frame, afr_nonblocking_entrylk_cbk,
(void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->fentrylk,
+ priv->children[index],
+ priv->children[index]->fops->fentrylk,
this->name, local->fd,
- basename,
+ int_lock->lockee[lockee_no].basename,
ENTRYLK_LOCK_NB, ENTRYLK_WRLCK,
NULL);
+ if (!--call_count)
+ break;
}
}
} else {
- GF_ASSERT (loc);
-
- call_count = internal_lock_count (frame, this);
+ call_count = int_lock->lockee_count * internal_lock_count (frame, this);
int_lock->lk_call_count = call_count;
int_lock->lk_expected_count = call_count;
- for (i = 0; i < priv->child_count; i++) {
- if (local->child_up[i]) {
- AFR_TRACE_ENTRYLK_IN (frame, this,
- AFR_ENTRYLK_NB_TRANSACTION,
- AFR_LOCK_OP, basename, i);
+ for (i = 0; i < int_lock->lockee_count*priv->child_count; i++) {
+ index = i%copies;
+ lockee_no = i/copies;
+ if (local->child_up[index]) {
+ AFR_TRACE_ENTRYLK_IN (frame, this, AFR_ENTRYLK_NB_TRANSACTION,
+ AFR_LOCK_OP,
+ int_lock->lockee[lockee_no].basename,
+ i);
STACK_WIND_COOKIE (frame, afr_nonblocking_entrylk_cbk,
(void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->entrylk,
- this->name, loc, basename,
+ priv->children[index],
+ priv->children[index]->fops->entrylk,
+ this->name, &int_lock->lockee[lockee_no].loc,
+ int_lock->lockee[lockee_no].basename,
ENTRYLK_LOCK_NB, ENTRYLK_WRLCK,
NULL);
if (!--call_count)
break;
-
}
}
}
@@ -1539,201 +1622,6 @@ out:
return ret;
}
-static int
-__is_lower_locked (call_frame_t *frame, xlator_t *this)
-{
- afr_internal_lock_t *int_lock = NULL;
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int count = 0;
- int i = 0;
-
- local = frame->local;
- int_lock = &local->internal_lock;
- priv = this->private;
-
- for (i = 0; i < priv->child_count; i++) {
- if (int_lock->lower_locked_nodes[i] & LOCKED_LOWER)
- count++;
- }
-
- return count;
-
-}
-
-static int
-__is_higher_locked (call_frame_t *frame, xlator_t *this)
-{
- afr_internal_lock_t *int_lock = NULL;
- afr_private_t *priv = NULL;
- afr_local_t *local = NULL;
- int count = 0;
- int i = 0;
-
- local = frame->local;
- int_lock = &local->internal_lock;
- priv = this->private;
-
- for (i = 0; i < priv->child_count; i++) {
- if (int_lock->locked_nodes[i] & LOCKED_YES)
- count++;
- }
-
- return count;
-
-}
-
-static int
-afr_unlock_lower_entrylk (call_frame_t *frame, xlator_t *this)
-{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- afr_private_t *priv = NULL;
- const char *basename = NULL;
- loc_t *loc = NULL;
- int call_count = 0;
- int i = -1;
-
- local = frame->local;
- int_lock = &local->internal_lock;
- priv = this->private;
-
- basename = int_lock->lk_basename;
- if (int_lock->lk_loc)
- loc = int_lock->lk_loc;
-
- call_count = __is_lower_locked (frame, this);
- int_lock->lk_call_count = call_count;
-
- if (!call_count){
- gf_log (this->name, GF_LOG_TRACE,
- "No internal locks unlocked");
- int_lock->lock_cbk (frame, this);
- goto out;
- }
-
- for (i = 0; i < priv->child_count; i++) {
- if (int_lock->lower_locked_nodes[i] & LOCKED_LOWER) {
- AFR_TRACE_ENTRYLK_IN (frame, this,
- AFR_ENTRYLK_NB_TRANSACTION,
- AFR_UNLOCK_OP, basename, i);
-
- STACK_WIND_COOKIE (frame, afr_unlock_entrylk_cbk,
- (void *) (long) i,
- priv->children[i],
- priv->children[i]->fops->entrylk,
- this->name,
- loc, basename,
- ENTRYLK_UNLOCK, ENTRYLK_WRLCK, NULL);
-
- if (!--call_count)
- break;
-
- }
- }
-
-out:
- return 0;
-
-}
-
-
-static int
-afr_post_unlock_higher_cbk (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t *local = NULL;
-
- local = frame->local;
-
- local->transaction.done (frame, this);
- return 0;
-}
-
-static int
-afr_post_unlock_lower_cbk (call_frame_t *frame, xlator_t *this)
-{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- loc_t *lower = NULL;
- loc_t *higher = NULL;
- const char *higher_name = NULL;
-
- local = frame->local;
- int_lock = &local->internal_lock;
-
- lower = lower_path (&local->transaction.parent_loc,
- local->transaction.basename,
- &local->transaction.new_parent_loc,
- local->transaction.new_basename);
-
- higher = (lower == &local->transaction.parent_loc ?
- &local->transaction.new_parent_loc :
- &local->transaction.parent_loc);
-
- higher_name = (higher == &local->transaction.parent_loc ?
- local->transaction.basename :
- local->transaction.new_basename);
-
- if (__is_higher_locked (frame, this)) {
- gf_log (this->name, GF_LOG_DEBUG,
- "unlocking higher");
- int_lock->lk_basename = higher_name;
- int_lock->lk_loc = higher;
- int_lock->lock_cbk = afr_post_unlock_higher_cbk;
-
- afr_unlock_entrylk (frame, this);
- } else
- local->transaction.done (frame, this);
-
- return 0;
-}
-
-static int
-afr_rename_unlock (call_frame_t *frame, xlator_t *this)
-{
- afr_internal_lock_t *int_lock = NULL;
- afr_local_t *local = NULL;
- loc_t *lower = NULL;
- const char *lower_name = NULL;
-
- local = frame->local;
- int_lock = &local->internal_lock;
-
- lower = lower_path (&local->transaction.parent_loc,
- local->transaction.basename,
- &local->transaction.new_parent_loc,
- local->transaction.new_basename);
-
- lower_name = (lower == &local->transaction.parent_loc ?
- local->transaction.basename :
- local->transaction.new_basename);
-
- if (__is_lower_locked (frame, this)) {
- gf_log (this->name, GF_LOG_DEBUG,
- "unlocking lower");
- int_lock->lk_basename = lower_name;
- int_lock->lk_loc = lower;
- int_lock->lock_cbk = afr_post_unlock_lower_cbk;
-
- afr_unlock_lower_entrylk (frame, this);
- } else
- afr_post_unlock_lower_cbk (frame, this);
-
- return 0;
-}
-
-static int
-afr_rename_transaction (call_frame_t *frame, xlator_t *this)
-{
- afr_local_t *local = NULL;
-
- local = frame->local;
-
- return (local->transaction.type ==
- AFR_ENTRY_RENAME_TRANSACTION);
-
-}
-
int32_t
afr_unlock (call_frame_t *frame, xlator_t *this)
{
@@ -1745,10 +1633,8 @@ afr_unlock (call_frame_t *frame, xlator_t *this)
if (is_afr_lock_transaction (local))
afr_unlock_inodelk (frame, this);
else
- if (!afr_rename_transaction (frame, this))
- afr_unlock_entrylk (frame, this);
- else
- afr_rename_unlock (frame, this);
+ afr_unlock_entrylk (frame, this);
+
} else {
if (is_afr_lock_selfheal (local))
afr_unlock_inodelk (frame, this);
diff --git a/xlators/cluster/afr/src/afr-self-heal-common.c b/xlators/cluster/afr/src/afr-self-heal-common.c
index f59a02557c1..9798aa4e0e6 100644
--- a/xlators/cluster/afr/src/afr-self-heal-common.c
+++ b/xlators/cluster/afr/src/afr-self-heal-common.c
@@ -1952,7 +1952,9 @@ afr_sh_entrylk (call_frame_t *frame, xlator_t *this, loc_t *loc,
{
afr_internal_lock_t *int_lock = NULL;
afr_local_t *local = NULL;
+ afr_private_t *priv = NULL;
+ priv = this->private;
local = frame->local;
int_lock = &local->internal_lock;
@@ -1965,6 +1967,10 @@ afr_sh_entrylk (call_frame_t *frame, xlator_t *this, loc_t *loc,
int_lock->lk_loc = loc;
int_lock->lock_cbk = lock_cbk;
+ int_lock->lockee_count = 0;
+ afr_init_entry_lockee (&int_lock->lockee[0], local, loc,
+ base_name, priv->child_count);
+ int_lock->lockee_count++;
afr_nonblocking_entrylk (frame, this);
return 0;
@@ -2023,6 +2029,7 @@ afr_local_t *afr_local_copy (afr_local_t *l, xlator_t *this)
afr_local_t *lc = NULL;
afr_self_heal_t *sh = NULL;
afr_self_heal_t *shc = NULL;
+ int i = 0;
priv = this->private;
@@ -2068,15 +2075,7 @@ afr_local_t *afr_local_copy (afr_local_t *l, xlator_t *this)
GF_CALLOC (sizeof (*l->internal_lock.inode_locked_nodes),
priv->child_count,
gf_afr_mt_char);
- if (l->internal_lock.entry_locked_nodes)
- lc->internal_lock.entry_locked_nodes =
- memdup (l->internal_lock.entry_locked_nodes,
- sizeof (*lc->internal_lock.entry_locked_nodes) * priv->child_count);
- else
- lc->internal_lock.entry_locked_nodes =
- GF_CALLOC (sizeof (*l->internal_lock.entry_locked_nodes),
- priv->child_count,
- gf_afr_mt_char);
+
if (l->internal_lock.locked_nodes)
lc->internal_lock.locked_nodes =
memdup (l->internal_lock.locked_nodes,
@@ -2087,11 +2086,38 @@ afr_local_t *afr_local_copy (afr_local_t *l, xlator_t *this)
priv->child_count,
gf_afr_mt_char);
+ for (i = 0; i < l->internal_lock.lockee_count; i++) {
+ loc_copy (&lc->internal_lock.lockee[i].loc,
+ &l->internal_lock.lockee[i].loc);
+
+ lc->internal_lock.lockee[i].locked_count =
+ l->internal_lock.lockee[i].locked_count;
+
+ if (l->internal_lock.lockee[i].basename)
+ lc->internal_lock.lockee[i].basename =
+ gf_strdup (l->internal_lock.lockee[i].basename);
+
+ if (l->internal_lock.lockee[i].locked_nodes) {
+ lc->internal_lock.lockee[i].locked_nodes =
+ memdup (l->internal_lock.lockee[i].locked_nodes,
+ sizeof (*lc->internal_lock.lockee[i].locked_nodes) *
+ priv->child_count);
+ } else {
+ lc->internal_lock.lockee[i].locked_nodes =
+ GF_CALLOC (priv->child_count,
+ sizeof (*lc->internal_lock.lockee[i].locked_nodes),
+ gf_afr_mt_char);
+ }
+
+ }
+ lc->internal_lock.lockee_count = l->internal_lock.lockee_count;
+
lc->internal_lock.inodelk_lock_count =
l->internal_lock.inodelk_lock_count;
lc->internal_lock.entrylk_lock_count =
l->internal_lock.entrylk_lock_count;
+
out:
return lc;
}
diff --git a/xlators/cluster/afr/src/afr-transaction.c b/xlators/cluster/afr/src/afr-transaction.c
index a1700783709..4e7ed55a0d9 100644
--- a/xlators/cluster/afr/src/afr-transaction.c
+++ b/xlators/cluster/afr/src/afr-transaction.c
@@ -466,7 +466,9 @@ afr_locked_nodes_get (afr_transaction_type type, afr_internal_lock_t *int_lock)
case AFR_ENTRY_TRANSACTION:
case AFR_ENTRY_RENAME_TRANSACTION:
- locked_nodes = int_lock->entry_locked_nodes;
+ /*Because same set of subvols participate in all lockee
+ * entities*/
+ locked_nodes = int_lock->lockee[0].locked_nodes;
break;
}
return locked_nodes;
@@ -1211,8 +1213,8 @@ afr_lock_rec (call_frame_t *frame, xlator_t *this)
case AFR_ENTRY_RENAME_TRANSACTION:
- int_lock->lock_cbk = afr_post_blocking_rename_cbk;
- afr_blocking_lock (frame, this);
+ int_lock->lock_cbk = afr_post_nonblocking_entrylk_cbk;
+ afr_nonblocking_entrylk (frame, this);
break;
case AFR_ENTRY_TRANSACTION:
diff --git a/xlators/cluster/afr/src/afr.h b/xlators/cluster/afr/src/afr.h
index a9c06370445..d31290fbf40 100644
--- a/xlators/cluster/afr/src/afr.h
+++ b/xlators/cluster/afr/src/afr.h
@@ -29,6 +29,8 @@
#define AFR_PATHINFO_HEADER "REPLICATE:"
#define AFR_SH_READDIR_SIZE_KEY "self-heal-readdir-size"
+#define AFR_LOCKEE_COUNT_MAX 3
+
struct _pump_private;
typedef int (*afr_expunge_done_cbk_t) (call_frame_t *frame, xlator_t *this,
@@ -342,11 +344,24 @@ afr_index_for_transaction_type (afr_transaction_type type)
return -1; /* make gcc happy */
}
+typedef struct {
+ loc_t loc;
+ char *basename;
+ unsigned char *locked_nodes;
+ int locked_count;
+
+} afr_entry_lockee_t;
+
+int
+afr_entry_lockee_cmp (const void *l1, const void *l2);
typedef struct {
loc_t *lk_loc;
struct gf_flock lk_flock;
+ int lockee_count;
+ afr_entry_lockee_t lockee[AFR_LOCKEE_COUNT_MAX];
+
const char *lk_basename;
const char *lower_basename;
const char *higher_basename;
@@ -356,7 +371,6 @@ typedef struct {
unsigned char *locked_nodes;
unsigned char *lower_locked_nodes;
unsigned char *inode_locked_nodes;
- unsigned char *entry_locked_nodes;
selfheal_lk_type_t selfheal_lk_type;
transaction_lk_type_t transaction_lk_type;
@@ -752,6 +766,13 @@ int32_t
afr_notify (xlator_t *this, int32_t event, void *data, void *data2);
int
+afr_init_entry_lockee (afr_entry_lockee_t *lockee, afr_local_t *local,
+ loc_t *loc, char *basename, int child_count);
+
+void
+afr_entry_lockee_cleanup (afr_internal_lock_t *int_lock);
+
+int
afr_attempt_lock_recovery (xlator_t *this, int32_t child_index);
int