Update Linux to v5.4.2
Change-Id: Idf6911045d9d382da2cfe01b1edff026404ac8fd
diff --git a/fs/ceph/snap.c b/fs/ceph/snap.c
index 041c27e..ccfcc66 100644
--- a/fs/ceph/snap.c
+++ b/fs/ceph/snap.c
@@ -3,12 +3,14 @@
#include <linux/sort.h>
#include <linux/slab.h>
-
+#include <linux/iversion.h>
#include "super.h"
#include "mds_client.h"
-
#include <linux/ceph/decode.h>
+/* unused map expires after 5 minutes */
+#define CEPH_SNAPID_MAP_TIMEOUT (5 * 60 * HZ)
+
/*
* Snapshots in ceph are driven in large part by cooperation from the
* client. In contrast to local file systems or file servers that
@@ -124,6 +126,8 @@
INIT_LIST_HEAD(&realm->inodes_with_caps);
spin_lock_init(&realm->inodes_with_caps_lock);
__insert_snap_realm(&mdsc->snap_realms, realm);
+ mdsc->num_snap_realms++;
+
dout("create_snap_realm %llx %p\n", realm->ino, realm);
return realm;
}
@@ -175,6 +179,7 @@
dout("__destroy_snap_realm %p %llx\n", realm, realm->ino);
rb_erase(&realm->node, &mdsc->snap_realms);
+ mdsc->num_snap_realms--;
if (realm->parent) {
list_del_init(&realm->child_item);
@@ -460,6 +465,7 @@
struct inode *inode = &ci->vfs_inode;
struct ceph_cap_snap *capsnap;
struct ceph_snap_context *old_snapc, *new_snapc;
+ struct ceph_buffer *old_blob = NULL;
int used, dirty;
capsnap = kzalloc(sizeof(*capsnap), GFP_NOFS);
@@ -536,7 +542,7 @@
capsnap->gid = inode->i_gid;
if (dirty & CEPH_CAP_XATTR_EXCL) {
- __ceph_build_xattrs_blob(ci);
+ old_blob = __ceph_build_xattrs_blob(ci);
capsnap->xattr_blob =
ceph_buffer_get(ci->i_xattrs.blob);
capsnap->xattr_version = ci->i_xattrs.version;
@@ -568,12 +574,18 @@
old_snapc = NULL;
update_snapc:
- if (ci->i_head_snapc) {
+ if (ci->i_wrbuffer_ref_head == 0 &&
+ ci->i_wr_ref == 0 &&
+ ci->i_dirty_caps == 0 &&
+ ci->i_flushing_caps == 0) {
+ ci->i_head_snapc = NULL;
+ } else {
ci->i_head_snapc = ceph_get_snap_context(new_snapc);
dout(" new snapc is %p\n", new_snapc);
}
spin_unlock(&ci->i_ceph_lock);
+ ceph_buffer_put(old_blob);
kfree(capsnap);
ceph_put_snap_context(old_snapc);
}
@@ -597,6 +609,8 @@
capsnap->mtime = inode->i_mtime;
capsnap->atime = inode->i_atime;
capsnap->ctime = inode->i_ctime;
+ capsnap->btime = ci->i_btime;
+ capsnap->change_attr = inode_peek_iversion_raw(inode);
capsnap->time_warp_seq = ci->i_time_warp_seq;
capsnap->truncate_size = ci->i_truncate_size;
capsnap->truncate_seq = ci->i_truncate_seq;
@@ -616,7 +630,8 @@
capsnap->size);
spin_lock(&mdsc->snap_flush_lock);
- list_add_tail(&ci->i_snap_flush_item, &mdsc->snap_flush_list);
+ if (list_empty(&ci->i_snap_flush_item))
+ list_add_tail(&ci->i_snap_flush_item, &mdsc->snap_flush_list);
spin_unlock(&mdsc->snap_flush_lock);
return 1; /* caller may want to ceph_flush_snaps */
}
@@ -638,13 +653,15 @@
if (!inode)
continue;
spin_unlock(&realm->inodes_with_caps_lock);
- iput(lastinode);
+ /* avoid calling iput_final() while holding
+ * mdsc->snap_rwsem or in mds dispatch threads */
+ ceph_async_iput(lastinode);
lastinode = inode;
ceph_queue_cap_snap(ci);
spin_lock(&realm->inodes_with_caps_lock);
}
spin_unlock(&realm->inodes_with_caps_lock);
- iput(lastinode);
+ ceph_async_iput(lastinode);
dout("queue_realm_cap_snaps %p %llx done\n", realm, realm->ino);
}
@@ -796,7 +813,9 @@
ihold(inode);
spin_unlock(&mdsc->snap_flush_lock);
ceph_flush_snaps(ci, &session);
- iput(inode);
+ /* avoid calling iput_final() while holding
+ * session->s_mutex or in mds dispatch threads */
+ ceph_async_iput(inode);
spin_lock(&mdsc->snap_flush_lock);
}
spin_unlock(&mdsc->snap_flush_lock);
@@ -940,12 +959,14 @@
ceph_get_snap_realm(mdsc, realm);
ceph_put_snap_realm(mdsc, oldrealm);
- iput(inode);
+ /* avoid calling iput_final() while holding
+ * mdsc->snap_rwsem or mds in dispatch threads */
+ ceph_async_iput(inode);
continue;
skip_inode:
spin_unlock(&ci->i_ceph_lock);
- iput(inode);
+ ceph_async_iput(inode);
}
/* we may have taken some of the old realm's children. */
@@ -985,3 +1006,154 @@
up_write(&mdsc->snap_rwsem);
return;
}
+
+struct ceph_snapid_map* ceph_get_snapid_map(struct ceph_mds_client *mdsc,
+ u64 snap)
+{
+ struct ceph_snapid_map *sm, *exist;
+ struct rb_node **p, *parent;
+ int ret;
+
+ exist = NULL;
+ spin_lock(&mdsc->snapid_map_lock);
+ p = &mdsc->snapid_map_tree.rb_node;
+ while (*p) {
+ exist = rb_entry(*p, struct ceph_snapid_map, node);
+ if (snap > exist->snap) {
+ p = &(*p)->rb_left;
+ } else if (snap < exist->snap) {
+ p = &(*p)->rb_right;
+ } else {
+ if (atomic_inc_return(&exist->ref) == 1)
+ list_del_init(&exist->lru);
+ break;
+ }
+ exist = NULL;
+ }
+ spin_unlock(&mdsc->snapid_map_lock);
+ if (exist) {
+ dout("found snapid map %llx -> %x\n", exist->snap, exist->dev);
+ return exist;
+ }
+
+ sm = kmalloc(sizeof(*sm), GFP_NOFS);
+ if (!sm)
+ return NULL;
+
+ ret = get_anon_bdev(&sm->dev);
+ if (ret < 0) {
+ kfree(sm);
+ return NULL;
+ }
+
+ INIT_LIST_HEAD(&sm->lru);
+ atomic_set(&sm->ref, 1);
+ sm->snap = snap;
+
+ exist = NULL;
+ parent = NULL;
+ p = &mdsc->snapid_map_tree.rb_node;
+ spin_lock(&mdsc->snapid_map_lock);
+ while (*p) {
+ parent = *p;
+ exist = rb_entry(*p, struct ceph_snapid_map, node);
+ if (snap > exist->snap)
+ p = &(*p)->rb_left;
+ else if (snap < exist->snap)
+ p = &(*p)->rb_right;
+ else
+ break;
+ exist = NULL;
+ }
+ if (exist) {
+ if (atomic_inc_return(&exist->ref) == 1)
+ list_del_init(&exist->lru);
+ } else {
+ rb_link_node(&sm->node, parent, p);
+ rb_insert_color(&sm->node, &mdsc->snapid_map_tree);
+ }
+ spin_unlock(&mdsc->snapid_map_lock);
+ if (exist) {
+ free_anon_bdev(sm->dev);
+ kfree(sm);
+ dout("found snapid map %llx -> %x\n", exist->snap, exist->dev);
+ return exist;
+ }
+
+ dout("create snapid map %llx -> %x\n", sm->snap, sm->dev);
+ return sm;
+}
+
+void ceph_put_snapid_map(struct ceph_mds_client* mdsc,
+ struct ceph_snapid_map *sm)
+{
+ if (!sm)
+ return;
+ if (atomic_dec_and_lock(&sm->ref, &mdsc->snapid_map_lock)) {
+ if (!RB_EMPTY_NODE(&sm->node)) {
+ sm->last_used = jiffies;
+ list_add_tail(&sm->lru, &mdsc->snapid_map_lru);
+ spin_unlock(&mdsc->snapid_map_lock);
+ } else {
+ /* already cleaned up by
+ * ceph_cleanup_snapid_map() */
+ spin_unlock(&mdsc->snapid_map_lock);
+ kfree(sm);
+ }
+ }
+}
+
+void ceph_trim_snapid_map(struct ceph_mds_client *mdsc)
+{
+ struct ceph_snapid_map *sm;
+ unsigned long now;
+ LIST_HEAD(to_free);
+
+ spin_lock(&mdsc->snapid_map_lock);
+ now = jiffies;
+
+ while (!list_empty(&mdsc->snapid_map_lru)) {
+ sm = list_first_entry(&mdsc->snapid_map_lru,
+ struct ceph_snapid_map, lru);
+ if (time_after(sm->last_used + CEPH_SNAPID_MAP_TIMEOUT, now))
+ break;
+
+ rb_erase(&sm->node, &mdsc->snapid_map_tree);
+ list_move(&sm->lru, &to_free);
+ }
+ spin_unlock(&mdsc->snapid_map_lock);
+
+ while (!list_empty(&to_free)) {
+ sm = list_first_entry(&to_free, struct ceph_snapid_map, lru);
+ list_del(&sm->lru);
+ dout("trim snapid map %llx -> %x\n", sm->snap, sm->dev);
+ free_anon_bdev(sm->dev);
+ kfree(sm);
+ }
+}
+
+void ceph_cleanup_snapid_map(struct ceph_mds_client *mdsc)
+{
+ struct ceph_snapid_map *sm;
+ struct rb_node *p;
+ LIST_HEAD(to_free);
+
+ spin_lock(&mdsc->snapid_map_lock);
+ while ((p = rb_first(&mdsc->snapid_map_tree))) {
+ sm = rb_entry(p, struct ceph_snapid_map, node);
+ rb_erase(p, &mdsc->snapid_map_tree);
+ RB_CLEAR_NODE(p);
+ list_move(&sm->lru, &to_free);
+ }
+ spin_unlock(&mdsc->snapid_map_lock);
+
+ while (!list_empty(&to_free)) {
+ sm = list_first_entry(&to_free, struct ceph_snapid_map, lru);
+ list_del(&sm->lru);
+ free_anon_bdev(sm->dev);
+ if (WARN_ON_ONCE(atomic_read(&sm->ref))) {
+ pr_err("snapid map %llx -> %x still in use\n",
+ sm->snap, sm->dev);
+ }
+ }
+}