struct ceph_cap_flush *ceph_alloc_cap_flush(void)
{
- return kmem_cache_alloc(ceph_cap_flush_cachep, GFP_KERNEL);
+ struct ceph_cap_flush *cf;
+
+ cf = kmem_cache_alloc(ceph_cap_flush_cachep, GFP_KERNEL);
+ cf->is_capsnap = false;
+ return cf;
}
void ceph_free_cap_flush(struct ceph_cap_flush *cf)
prev->wake = true;
wake = false;
}
- list_del(&cf->g_list);
+ list_del_init(&cf->g_list);
return wake;
}
prev->wake = true;
wake = false;
}
- list_del(&cf->i_list);
+ list_del_init(&cf->i_list);
return wake;
}
*/
static int unsafe_request_wait(struct inode *inode)
{
+ struct ceph_mds_client *mdsc = ceph_sb_to_client(inode->i_sb)->mdsc;
struct ceph_inode_info *ci = ceph_inode(inode);
struct ceph_mds_request *req1 = NULL, *req2 = NULL;
int ret, err = 0;
}
spin_unlock(&ci->i_unsafe_lock);
+ /*
+ * Trigger to flush the journal logs in all the relevant MDSes
+ * manually, or in the worst case we must wait at most 5 seconds
+ * to wait the journal logs to be flushed by the MDSes periodically.
+ */
+ if (req1 || req2) {
+ struct ceph_mds_session **sessions = NULL;
+ struct ceph_mds_session *s;
+ struct ceph_mds_request *req;
+ unsigned int max;
+ int i;
+
+ /*
+ * The mdsc->max_sessions is unlikely to be changed
+ * mostly, here we will retry it by reallocating the
+ * sessions arrary memory to get rid of the mdsc->mutex
+ * lock.
+ */
+retry:
+ max = mdsc->max_sessions;
+ sessions = krealloc(sessions, max * sizeof(s), __GFP_ZERO);
+ if (!sessions)
+ return -ENOMEM;
+
+ spin_lock(&ci->i_unsafe_lock);
+ if (req1) {
+ list_for_each_entry(req, &ci->i_unsafe_dirops,
+ r_unsafe_dir_item) {
+ s = req->r_session;
+ if (unlikely(s->s_mds > max)) {
+ spin_unlock(&ci->i_unsafe_lock);
+ goto retry;
+ }
+ if (!sessions[s->s_mds]) {
+ s = ceph_get_mds_session(s);
+ sessions[s->s_mds] = s;
+ }
+ }
+ }
+ if (req2) {
+ list_for_each_entry(req, &ci->i_unsafe_iops,
+ r_unsafe_target_item) {
+ s = req->r_session;
+ if (unlikely(s->s_mds > max)) {
+ spin_unlock(&ci->i_unsafe_lock);
+ goto retry;
+ }
+ if (!sessions[s->s_mds]) {
+ s = ceph_get_mds_session(s);
+ sessions[s->s_mds] = s;
+ }
+ }
+ }
+ spin_unlock(&ci->i_unsafe_lock);
+
+ /* the auth MDS */
+ spin_lock(&ci->i_ceph_lock);
+ if (ci->i_auth_cap) {
+ s = ci->i_auth_cap->session;
+ if (!sessions[s->s_mds])
+ sessions[s->s_mds] = ceph_get_mds_session(s);
+ }
+ spin_unlock(&ci->i_ceph_lock);
+
+ /* send flush mdlog request to MDSes */
+ for (i = 0; i < max; i++) {
+ s = sessions[i];
+ if (s) {
+ send_flush_mdlog(s);
+ ceph_put_mds_session(s);
+ }
+ }
+ kfree(sessions);
+ }
+
dout("unsafe_request_wait %p wait on tid %llu %llu\n",
inode, req1 ? req1->r_tid : 0ULL, req2 ? req2->r_tid : 0ULL);
if (req1) {
ci->i_ceph_flags &= ~CEPH_I_KICK_FLUSH;
list_for_each_entry_reverse(cf, &ci->i_cap_flush_list, i_list) {
- if (!cf->caps) {
+ if (cf->is_capsnap) {
last_snap_flush = cf->tid;
break;
}
first_tid = cf->tid + 1;
- if (cf->caps) {
+ if (!cf->is_capsnap) {
struct cap_msg_args arg;
dout("kick_flushing_caps %p cap %p tid %llu %s\n",
cleaned = cf->caps;
/* Is this a capsnap? */
- if (cf->caps == 0)
+ if (cf->is_capsnap)
continue;
if (cf->tid <= flush_tid) {
while (!list_empty(&to_remove)) {
cf = list_first_entry(&to_remove,
struct ceph_cap_flush, i_list);
- list_del(&cf->i_list);
- ceph_free_cap_flush(cf);
+ list_del_init(&cf->i_list);
+ if (!cf->is_capsnap)
+ ceph_free_cap_flush(cf);
}
if (wake_ci)
done:
mutex_unlock(&session->s_mutex);
done_unlocked:
- ceph_put_string(extra_info.pool_ns);
iput(inode);
+out:
+ ceph_put_string(extra_info.pool_ns);
return;
flush_cap_releases:
bad:
pr_err("ceph_handle_caps: corrupt message\n");
ceph_msg_dump(msg);
- return;
+ goto out;
}
/*
* Delayed work handler to process end of delayed cap release LRU list.
+ *
+ * If new caps are added to the list while processing it, these won't get
+ * processed in this run. In this case, the ci->i_hold_caps_max will be
+ * returned so that the work can be scheduled accordingly.
*/
-void ceph_check_delayed_caps(struct ceph_mds_client *mdsc)
+unsigned long ceph_check_delayed_caps(struct ceph_mds_client *mdsc)
{
struct inode *inode;
struct ceph_inode_info *ci;
+ struct ceph_mount_options *opt = mdsc->fsc->mount_options;
+ unsigned long delay_max = opt->caps_wanted_delay_max * HZ;
+ unsigned long loop_start = jiffies;
+ unsigned long delay = 0;
dout("check_delayed_caps\n");
spin_lock(&mdsc->cap_delay_lock);
ci = list_first_entry(&mdsc->cap_delay_list,
struct ceph_inode_info,
i_cap_delay_list);
+ if (time_before(loop_start, ci->i_hold_caps_max - delay_max)) {
+ dout("%s caps added recently. Exiting loop", __func__);
+ delay = ci->i_hold_caps_max;
+ break;
+ }
if ((ci->i_ceph_flags & CEPH_I_FLUSH) == 0 &&
time_before(jiffies, ci->i_hold_caps_max))
break;
}
}
spin_unlock(&mdsc->cap_delay_lock);
+
+ return delay;
}
/*
dout("flush_dirty_caps done\n");
}
-static void iterate_sessions(struct ceph_mds_client *mdsc,
- void (*cb)(struct ceph_mds_session *))
-{
- int mds;
-
- mutex_lock(&mdsc->mutex);
- for (mds = 0; mds < mdsc->max_sessions; ++mds) {
- struct ceph_mds_session *s;
-
- if (!mdsc->sessions[mds])
- continue;
-
- s = ceph_get_mds_session(mdsc->sessions[mds]);
- if (!s)
- continue;
-
- mutex_unlock(&mdsc->mutex);
- cb(s);
- ceph_put_mds_session(s);
- mutex_lock(&mdsc->mutex);
- }
- mutex_unlock(&mdsc->mutex);
-}
-
void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc)
{
- iterate_sessions(mdsc, flush_dirty_session_caps);
+ ceph_mdsc_iterate_sessions(mdsc, flush_dirty_session_caps, true);
}
void __ceph_touch_fmode(struct ceph_inode_info *ci,
projects / platform / kernel / linux-rpi.git / blobdiff