summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorYan, Zheng <zyan@redhat.com>2015-05-01 17:49:16 +0800
committerIlya Dryomov <idryomov@gmail.com>2015-06-25 11:49:28 +0300
commit604d1b0245b97738cde4341944ad93edff4b2827 (patch)
tree05a75dfc1ddf4729405f9cb6ce2f405b8e47791e
parent860560904962d08fd38666207c910065fe53e074 (diff)
downloadop-kernel-dev-604d1b0245b97738cde4341944ad93edff4b2827.zip
op-kernel-dev-604d1b0245b97738cde4341944ad93edff4b2827.tar.gz
ceph: take snap_rwsem when accessing snap realm's cached_context
When ceph inode's i_head_snapc is NULL, __ceph_mark_dirty_caps() accesses snap realm's cached_context. So we need take read lock of snap_rwsem. Signed-off-by: Yan, Zheng <zyan@redhat.com>
-rw-r--r--fs/ceph/caps.c4
-rw-r--r--fs/ceph/inode.c15
-rw-r--r--fs/ceph/xattr.c45
3 files changed, 57 insertions, 7 deletions
diff --git a/fs/ceph/caps.c b/fs/ceph/caps.c
index f1dbcae..900c05f 100644
--- a/fs/ceph/caps.c
+++ b/fs/ceph/caps.c
@@ -1413,9 +1413,11 @@ int __ceph_mark_dirty_caps(struct ceph_inode_info *ci, int mask)
ceph_cap_string(was | mask));
ci->i_dirty_caps |= mask;
if (was == 0) {
- if (!ci->i_head_snapc)
+ if (!ci->i_head_snapc) {
+ WARN_ON_ONCE(!rwsem_is_locked(&mdsc->snap_rwsem));
ci->i_head_snapc = ceph_get_snap_context(
ci->i_snap_realm->cached_context);
+ }
dout(" inode %p now dirty snapc %p auth cap %p\n",
&ci->vfs_inode, ci->i_head_snapc, ci->i_auth_cap);
BUG_ON(!list_empty(&ci->i_dirty_item));
diff --git a/fs/ceph/inode.c b/fs/ceph/inode.c
index 1a68c0e..1c991df 100644
--- a/fs/ceph/inode.c
+++ b/fs/ceph/inode.c
@@ -1727,6 +1727,7 @@ int ceph_setattr(struct dentry *dentry, struct iattr *attr)
int mask = 0;
int err = 0;
int inode_dirty_flags = 0;
+ bool lock_snap_rwsem = false;
if (ceph_snap(inode) != CEPH_NOSNAP)
return -EROFS;
@@ -1742,6 +1743,18 @@ int ceph_setattr(struct dentry *dentry, struct iattr *attr)
spin_lock(&ci->i_ceph_lock);
issued = __ceph_caps_issued(ci, NULL);
+
+ if (!ci->i_head_snapc &&
+ (issued & (CEPH_CAP_ANY_EXCL | CEPH_CAP_FILE_WR))) {
+ lock_snap_rwsem = true;
+ if (!down_read_trylock(&mdsc->snap_rwsem)) {
+ spin_unlock(&ci->i_ceph_lock);
+ down_read(&mdsc->snap_rwsem);
+ spin_lock(&ci->i_ceph_lock);
+ issued = __ceph_caps_issued(ci, NULL);
+ }
+ }
+
dout("setattr %p issued %s\n", inode, ceph_cap_string(issued));
if (ia_valid & ATTR_UID) {
@@ -1890,6 +1903,8 @@ int ceph_setattr(struct dentry *dentry, struct iattr *attr)
release &= issued;
spin_unlock(&ci->i_ceph_lock);
+ if (lock_snap_rwsem)
+ up_read(&mdsc->snap_rwsem);
if (inode_dirty_flags)
__mark_inode_dirty(inode, inode_dirty_flags);
diff --git a/fs/ceph/xattr.c b/fs/ceph/xattr.c
index cd7ffad4..c6f7d9b 100644
--- a/fs/ceph/xattr.c
+++ b/fs/ceph/xattr.c
@@ -911,6 +911,7 @@ int __ceph_setxattr(struct dentry *dentry, const char *name,
struct inode *inode = d_inode(dentry);
struct ceph_vxattr *vxattr;
struct ceph_inode_info *ci = ceph_inode(inode);
+ struct ceph_mds_client *mdsc = ceph_sb_to_client(dentry->d_sb)->mdsc;
int issued;
int err;
int dirty = 0;
@@ -920,6 +921,7 @@ int __ceph_setxattr(struct dentry *dentry, const char *name,
char *newval = NULL;
struct ceph_inode_xattr *xattr = NULL;
int required_blob_size;
+ bool lock_snap_rwsem = false;
if (!ceph_is_valid_xattr(name))
return -EOPNOTSUPP;
@@ -951,9 +953,20 @@ int __ceph_setxattr(struct dentry *dentry, const char *name,
spin_lock(&ci->i_ceph_lock);
retry:
issued = __ceph_caps_issued(ci, NULL);
- dout("setxattr %p issued %s\n", inode, ceph_cap_string(issued));
if (ci->i_xattrs.version == 0 || !(issued & CEPH_CAP_XATTR_EXCL))
goto do_sync;
+
+ if (!lock_snap_rwsem && !ci->i_head_snapc) {
+ lock_snap_rwsem = true;
+ if (!down_read_trylock(&mdsc->snap_rwsem)) {
+ spin_unlock(&ci->i_ceph_lock);
+ down_read(&mdsc->snap_rwsem);
+ spin_lock(&ci->i_ceph_lock);
+ goto retry;
+ }
+ }
+
+ dout("setxattr %p issued %s\n", inode, ceph_cap_string(issued));
__build_xattrs(inode);
required_blob_size = __get_required_blob_size(ci, name_len, val_len);
@@ -966,7 +979,7 @@ retry:
dout(" preaallocating new blob size=%d\n", required_blob_size);
blob = ceph_buffer_new(required_blob_size, GFP_NOFS);
if (!blob)
- goto out;
+ goto do_sync_unlocked;
spin_lock(&ci->i_ceph_lock);
if (ci->i_xattrs.prealloc_blob)
ceph_buffer_put(ci->i_xattrs.prealloc_blob);
@@ -984,6 +997,8 @@ retry:
}
spin_unlock(&ci->i_ceph_lock);
+ if (lock_snap_rwsem)
+ up_read(&mdsc->snap_rwsem);
if (dirty)
__mark_inode_dirty(inode, dirty);
return err;
@@ -991,6 +1006,8 @@ retry:
do_sync:
spin_unlock(&ci->i_ceph_lock);
do_sync_unlocked:
+ if (lock_snap_rwsem)
+ up_read(&mdsc->snap_rwsem);
err = ceph_sync_setxattr(dentry, name, value, size, flags);
out:
kfree(newname);
@@ -1044,10 +1061,12 @@ int __ceph_removexattr(struct dentry *dentry, const char *name)
struct inode *inode = d_inode(dentry);
struct ceph_vxattr *vxattr;
struct ceph_inode_info *ci = ceph_inode(inode);
+ struct ceph_mds_client *mdsc = ceph_sb_to_client(dentry->d_sb)->mdsc;
int issued;
int err;
int required_blob_size;
int dirty;
+ bool lock_snap_rwsem = false;
if (!ceph_is_valid_xattr(name))
return -EOPNOTSUPP;
@@ -1064,10 +1083,21 @@ int __ceph_removexattr(struct dentry *dentry, const char *name)
spin_lock(&ci->i_ceph_lock);
retry:
issued = __ceph_caps_issued(ci, NULL);
- dout("removexattr %p issued %s\n", inode, ceph_cap_string(issued));
-
if (ci->i_xattrs.version == 0 || !(issued & CEPH_CAP_XATTR_EXCL))
goto do_sync;
+
+ if (!lock_snap_rwsem && !ci->i_head_snapc) {
+ lock_snap_rwsem = true;
+ if (!down_read_trylock(&mdsc->snap_rwsem)) {
+ spin_unlock(&ci->i_ceph_lock);
+ down_read(&mdsc->snap_rwsem);
+ spin_lock(&ci->i_ceph_lock);
+ goto retry;
+ }
+ }
+
+ dout("removexattr %p issued %s\n", inode, ceph_cap_string(issued));
+
__build_xattrs(inode);
required_blob_size = __get_required_blob_size(ci, 0, 0);
@@ -1080,7 +1110,7 @@ retry:
dout(" preaallocating new blob size=%d\n", required_blob_size);
blob = ceph_buffer_new(required_blob_size, GFP_NOFS);
if (!blob)
- goto out;
+ goto do_sync_unlocked;
spin_lock(&ci->i_ceph_lock);
if (ci->i_xattrs.prealloc_blob)
ceph_buffer_put(ci->i_xattrs.prealloc_blob);
@@ -1094,14 +1124,17 @@ retry:
ci->i_xattrs.dirty = true;
inode->i_ctime = CURRENT_TIME;
spin_unlock(&ci->i_ceph_lock);
+ if (lock_snap_rwsem)
+ up_read(&mdsc->snap_rwsem);
if (dirty)
__mark_inode_dirty(inode, dirty);
return err;
do_sync:
spin_unlock(&ci->i_ceph_lock);
do_sync_unlocked:
+ if (lock_snap_rwsem)
+ up_read(&mdsc->snap_rwsem);
err = ceph_send_removexattr(dentry, name);
-out:
return err;
}
OpenPOWER on IntegriCloud