summaryrefslogtreecommitdiff
path: root/fs/ceph/caps.c
diff options
context:
space:
mode:
Diffstat (limited to 'fs/ceph/caps.c')
-rw-r--r--fs/ceph/caps.c82
1 files changed, 54 insertions, 28 deletions
diff --git a/fs/ceph/caps.c b/fs/ceph/caps.c
index f1ad6884d4da..bf2e94005598 100644
--- a/fs/ceph/caps.c
+++ b/fs/ceph/caps.c
@@ -1577,7 +1577,7 @@ static void __ceph_flush_snaps(struct ceph_inode_info *ci,
while (first_tid <= last_tid) {
struct ceph_cap *cap = ci->i_auth_cap;
- struct ceph_cap_flush *cf;
+ struct ceph_cap_flush *cf = NULL, *iter;
int ret;
if (!(cap && cap->session == session)) {
@@ -1587,8 +1587,9 @@ static void __ceph_flush_snaps(struct ceph_inode_info *ci,
}
ret = -ENOENT;
- list_for_each_entry(cf, &ci->i_cap_flush_list, i_list) {
- if (cf->tid >= first_tid) {
+ list_for_each_entry(iter, &ci->i_cap_flush_list, i_list) {
+ if (iter->tid >= first_tid) {
+ cf = iter;
ret = 0;
break;
}
@@ -1910,6 +1911,7 @@ void ceph_check_caps(struct ceph_inode_info *ci, int flags,
struct rb_node *p;
bool queue_invalidate = false;
bool tried_invalidate = false;
+ bool queue_writeback = false;
if (session)
ceph_get_mds_session(session);
@@ -2062,10 +2064,27 @@ retry:
}
/* completed revocation? going down and there are no caps? */
- if (revoking && (revoking & cap_used) == 0) {
- dout("completed revocation of %s\n",
- ceph_cap_string(cap->implemented & ~cap->issued));
- goto ack;
+ if (revoking) {
+ if ((revoking & cap_used) == 0) {
+ dout("completed revocation of %s\n",
+ ceph_cap_string(cap->implemented & ~cap->issued));
+ goto ack;
+ }
+
+ /*
+ * If the "i_wrbuffer_ref" was increased by mmap or generic
+ * cache write just before the ceph_check_caps() is called,
+ * the Fb capability revoking will fail this time. Then we
+ * must wait for the BDI's delayed work to flush the dirty
+ * pages and to release the "i_wrbuffer_ref", which will cost
+ * at most 5 seconds. That means the MDS needs to wait at
+ * most 5 seconds to finished the Fb capability's revocation.
+ *
+ * Let's queue a writeback for it.
+ */
+ if (S_ISREG(inode->i_mode) && ci->i_wrbuffer_ref &&
+ (revoking & CEPH_CAP_FILE_BUFFER))
+ queue_writeback = true;
}
/* want more caps from mds? */
@@ -2135,6 +2154,8 @@ ack:
spin_unlock(&ci->i_ceph_lock);
ceph_put_mds_session(session);
+ if (queue_writeback)
+ ceph_queue_writeback(inode);
if (queue_invalidate)
ceph_queue_invalidate(inode);
}
@@ -2218,9 +2239,9 @@ static int caps_are_flushed(struct inode *inode, u64 flush_tid)
}
/*
- * wait for any unsafe requests to complete.
+ * flush the mdlog and wait for any unsafe requests to complete.
*/
-static int unsafe_request_wait(struct inode *inode)
+static int flush_mdlog_and_wait_inode_unsafe_requests(struct inode *inode)
{
struct ceph_mds_client *mdsc = ceph_sb_to_client(inode->i_sb)->mdsc;
struct ceph_inode_info *ci = ceph_inode(inode);
@@ -2274,6 +2295,8 @@ retry:
list_for_each_entry(req, &ci->i_unsafe_dirops,
r_unsafe_dir_item) {
s = req->r_session;
+ if (!s)
+ continue;
if (unlikely(s->s_mds >= max_sessions)) {
spin_unlock(&ci->i_unsafe_lock);
for (i = 0; i < max_sessions; i++) {
@@ -2294,6 +2317,8 @@ retry:
list_for_each_entry(req, &ci->i_unsafe_iops,
r_unsafe_target_item) {
s = req->r_session;
+ if (!s)
+ continue;
if (unlikely(s->s_mds >= max_sessions)) {
spin_unlock(&ci->i_unsafe_lock);
for (i = 0; i < max_sessions; i++) {
@@ -2332,7 +2357,7 @@ retry:
kfree(sessions);
}
- dout("unsafe_request_wait %p wait on tid %llu %llu\n",
+ dout("%s %p wait on tid %llu %llu\n", __func__,
inode, req1 ? req1->r_tid : 0ULL, req2 ? req2->r_tid : 0ULL);
if (req1) {
ret = !wait_for_completion_timeout(&req1->r_safe_completion,
@@ -2376,7 +2401,7 @@ int ceph_fsync(struct file *file, loff_t start, loff_t end, int datasync)
dirty = try_flush_caps(inode, &flush_tid);
dout("fsync dirty caps are %s\n", ceph_cap_string(dirty));
- err = unsafe_request_wait(inode);
+ err = flush_mdlog_and_wait_inode_unsafe_requests(inode);
/*
* only wait on non-file metadata writeback (the mds
@@ -3178,10 +3203,9 @@ void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
struct ceph_snap_context *snapc)
{
struct inode *inode = &ci->vfs_inode;
- struct ceph_cap_snap *capsnap = NULL;
+ struct ceph_cap_snap *capsnap = NULL, *iter;
int put = 0;
bool last = false;
- bool found = false;
bool flush_snaps = false;
bool complete_capsnap = false;
@@ -3208,14 +3232,14 @@ void ceph_put_wrbuffer_cap_refs(struct ceph_inode_info *ci, int nr,
ci->i_wrbuffer_ref, ci->i_wrbuffer_ref_head,
last ? " LAST" : "");
} else {
- list_for_each_entry(capsnap, &ci->i_cap_snaps, ci_item) {
- if (capsnap->context == snapc) {
- found = true;
+ list_for_each_entry(iter, &ci->i_cap_snaps, ci_item) {
+ if (iter->context == snapc) {
+ capsnap = iter;
break;
}
}
- if (!found) {
+ if (!capsnap) {
/*
* The capsnap should already be removed when removing
* auth cap in the case of a forced unmount.
@@ -3765,8 +3789,7 @@ static void handle_cap_flushsnap_ack(struct inode *inode, u64 flush_tid,
struct ceph_inode_info *ci = ceph_inode(inode);
struct ceph_mds_client *mdsc = ceph_sb_to_client(inode->i_sb)->mdsc;
u64 follows = le64_to_cpu(m->snap_follows);
- struct ceph_cap_snap *capsnap;
- bool flushed = false;
+ struct ceph_cap_snap *capsnap = NULL, *iter;
bool wake_ci = false;
bool wake_mdsc = false;
@@ -3774,26 +3797,26 @@ static void handle_cap_flushsnap_ack(struct inode *inode, u64 flush_tid,
inode, ci, session->s_mds, follows);
spin_lock(&ci->i_ceph_lock);
- list_for_each_entry(capsnap, &ci->i_cap_snaps, ci_item) {
- if (capsnap->follows == follows) {
- if (capsnap->cap_flush.tid != flush_tid) {
+ list_for_each_entry(iter, &ci->i_cap_snaps, ci_item) {
+ if (iter->follows == follows) {
+ if (iter->cap_flush.tid != flush_tid) {
dout(" cap_snap %p follows %lld tid %lld !="
- " %lld\n", capsnap, follows,
- flush_tid, capsnap->cap_flush.tid);
+ " %lld\n", iter, follows,
+ flush_tid, iter->cap_flush.tid);
break;
}
- flushed = true;
+ capsnap = iter;
break;
} else {
dout(" skipping cap_snap %p follows %lld\n",
- capsnap, capsnap->follows);
+ iter, iter->follows);
}
}
- if (flushed)
+ if (capsnap)
ceph_remove_capsnap(inode, capsnap, &wake_ci, &wake_mdsc);
spin_unlock(&ci->i_ceph_lock);
- if (flushed) {
+ if (capsnap) {
ceph_put_snap_context(capsnap->context);
ceph_put_cap_snap(capsnap);
if (wake_ci)
@@ -3870,6 +3893,7 @@ static void handle_cap_export(struct inode *inode, struct ceph_mds_caps *ex,
dout("handle_cap_export inode %p ci %p mds%d mseq %d target %d\n",
inode, ci, mds, mseq, target);
retry:
+ down_read(&mdsc->snap_rwsem);
spin_lock(&ci->i_ceph_lock);
cap = __get_cap_for_mds(ci, mds);
if (!cap || cap->cap_id != le64_to_cpu(ex->cap_id))
@@ -3933,6 +3957,7 @@ retry:
}
spin_unlock(&ci->i_ceph_lock);
+ up_read(&mdsc->snap_rwsem);
mutex_unlock(&session->s_mutex);
/* open target session */
@@ -3958,6 +3983,7 @@ retry:
out_unlock:
spin_unlock(&ci->i_ceph_lock);
+ up_read(&mdsc->snap_rwsem);
mutex_unlock(&session->s_mutex);
if (tsession) {
mutex_unlock(&tsession->s_mutex);
generated by cgit v1.2.3 (git 2.39.0) at 2025-01-08 02:21:37 +0300