libceph: a major OSD client update

This is a major sync up, up to ~Jewel.  The highlights are:

- per-session request trees (vs a global per-client tree)
- per-session locking (vs a global per-client rwlock)
- homeless OSD session
- no ad-hoc global per-client lists
- support for pool quotas
- foundation for watch/notify v2 support
- foundation for map check (pool deletion detection) support

The switchover is incomplete: lingering requests can be setup and
teared down but aren't ever reestablished.  This functionality is
restored with the introduction of the new lingering infrastructure
(ceph_osd_linger_request, linger_work, etc) in a later commit.
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>

fs/ceph/ioctl.c

@@ -193,12 +193,12 @@ static long ceph_ioctl_get_dataloc(struct file *file, void __user *arg)
 	if (copy_from_user(&dl, arg, sizeof(dl)))
 		return -EFAULT;
 
-	down_read(&osdc->map_sem);
+	down_read(&osdc->lock);
 	r = ceph_calc_file_object_mapping(&ci->i_layout, dl.file_offset, len,
 					  &dl.object_no, &dl.object_offset,
 					  &olen);
 	if (r < 0) {
-		up_read(&osdc->map_sem);
+		up_read(&osdc->lock);
 		return -EIO;
 	}
 	dl.file_offset -= dl.object_offset;
@@ -217,7 +217,7 @@ static long ceph_ioctl_get_dataloc(struct file *file, void __user *arg)
 
 	r = ceph_object_locator_to_pg(osdc->osdmap, &oid, &oloc, &pgid);
 	if (r < 0) {
-		up_read(&osdc->map_sem);
+		up_read(&osdc->lock);
 		return r;
 	}
 
@@ -230,7 +230,7 @@ static long ceph_ioctl_get_dataloc(struct file *file, void __user *arg)
 	} else {
 		memset(&dl.osd_addr, 0, sizeof(dl.osd_addr));
 	}
-	up_read(&osdc->map_sem);
+	up_read(&osdc->lock);
 
 	/* send result back to user */
 	if (copy_to_user(arg, &dl, sizeof(dl)))

fs/ceph/xattr.c

@@ -75,7 +75,7 @@ static size_t ceph_vxattrcb_layout(struct ceph_inode_info *ci, char *val,
 	char buf[128];
 
 	dout("ceph_vxattrcb_layout %p\n", &ci->vfs_inode);
-	down_read(&osdc->map_sem);
+	down_read(&osdc->lock);
 	pool_name = ceph_pg_pool_name_by_id(osdc->osdmap, pool);
 	if (pool_name) {
 		size_t len = strlen(pool_name);
@@ -107,7 +107,7 @@ static size_t ceph_vxattrcb_layout(struct ceph_inode_info *ci, char *val,
 				ret = -ERANGE;
 		}
 	}
-	up_read(&osdc->map_sem);
+	up_read(&osdc->lock);
 	return ret;
 }
 
@@ -141,13 +141,13 @@ static size_t ceph_vxattrcb_layout_pool(struct ceph_inode_info *ci,
 	s64 pool = ceph_file_layout_pg_pool(ci->i_layout);
 	const char *pool_name;
 
-	down_read(&osdc->map_sem);
+	down_read(&osdc->lock);
 	pool_name = ceph_pg_pool_name_by_id(osdc->osdmap, pool);
 	if (pool_name)
 		ret = snprintf(val, size, "%s", pool_name);
 	else
 		ret = snprintf(val, size, "%lld", (unsigned long long)pool);
-	up_read(&osdc->map_sem);
+	up_read(&osdc->lock);
 	return ret;
 }
 

include/linux/ceph/osd_client.h

@@ -33,12 +33,13 @@ struct ceph_osd {
 	int o_incarnation;
 	struct rb_node o_node;
 	struct ceph_connection o_con;
-	struct list_head o_requests;
+	struct rb_root o_requests;
 	struct list_head o_linger_requests;
 	struct list_head o_osd_lru;
 	struct ceph_auth_handshake o_auth;
 	unsigned long lru_ttl;
 	struct list_head o_keepalive_item;
+	struct mutex lock;
 };
 
 #define CEPH_OSD_SLAB_OPS	2
@@ -144,8 +145,6 @@ struct ceph_osd_request_target {
 struct ceph_osd_request {
 	u64             r_tid;              /* unique for this client */
 	struct rb_node  r_node;
-	struct list_head r_req_lru_item;
-	struct list_head r_osd_item;
 	struct list_head r_linger_item;
 	struct list_head r_linger_osd_item;
 	struct ceph_osd *r_osd;
@@ -219,19 +218,16 @@ struct ceph_osd_client {
 	struct ceph_client     *client;
 
 	struct ceph_osdmap     *osdmap;       /* current map */
-	struct rw_semaphore    map_sem;
+	struct rw_semaphore    lock;
 
-	struct mutex           request_mutex;
 	struct rb_root         osds;          /* osds */
 	struct list_head       osd_lru;       /* idle osds */
 	spinlock_t             osd_lru_lock;
-	u64                    last_tid;      /* tid of last request */
-	struct rb_root         requests;      /* pending requests */
-	struct list_head       req_lru;	      /* in-flight lru */
-	struct list_head       req_unsent;    /* unsent/need-resend queue */
-	struct list_head       req_notarget;  /* map to no osd */
 	struct list_head       req_linger;    /* lingering requests */
-	int                    num_requests;
+	struct ceph_osd        homeless_osd;
+	atomic64_t             last_tid;      /* tid of last request */
+	atomic_t               num_requests;
+	atomic_t               num_homeless;
 	struct delayed_work    timeout_work;
 	struct delayed_work    osds_timeout_work;
 #ifdef CONFIG_DEBUG_FS

net/ceph/debugfs.c

@@ -182,21 +182,39 @@ static void dump_request(struct seq_file *s, struct ceph_osd_request *req)
 	seq_putc(s, '\n');
 }
 
+static void dump_requests(struct seq_file *s, struct ceph_osd *osd)
+{
+	struct rb_node *n;
+
+	mutex_lock(&osd->lock);
+	for (n = rb_first(&osd->o_requests); n; n = rb_next(n)) {
+		struct ceph_osd_request *req =
+		    rb_entry(n, struct ceph_osd_request, r_node);
+
+		dump_request(s, req);
+	}
+
+	mutex_unlock(&osd->lock);
+}
+
 static int osdc_show(struct seq_file *s, void *pp)
 {
 	struct ceph_client *client = s->private;
 	struct ceph_osd_client *osdc = &client->osdc;
-	struct rb_node *p;
-
-	mutex_lock(&osdc->request_mutex);
-	for (p = rb_first(&osdc->requests); p; p = rb_next(p)) {
-		struct ceph_osd_request *req;
+	struct rb_node *n;
 
-		req = rb_entry(p, struct ceph_osd_request, r_node);
+	down_read(&osdc->lock);
+	seq_printf(s, "REQUESTS %d homeless %d\n",
+		   atomic_read(&osdc->num_requests),
+		   atomic_read(&osdc->num_homeless));
+	for (n = rb_first(&osdc->osds); n; n = rb_next(n)) {
+		struct ceph_osd *osd = rb_entry(n, struct ceph_osd, o_node);
 
-		dump_request(s, req);
+		dump_requests(s, osd);
 	}
-	mutex_unlock(&osdc->request_mutex);
+	dump_requests(s, &osdc->homeless_osd);
+
+	up_read(&osdc->lock);
 	return 0;
 }