libceph: don't consume a ref on pagelist in ceph_msg_data_add_pagelist()

Because send_mds_reconnect() wants to send a message with a pagelist
and pass the ownership to the messenger, ceph_msg_data_add_pagelist()
consumes a ref which is then put in ceph_msg_data_destroy().  This
makes managing pagelists in the OSD client (where they are wrapped in
ceph_osd_data) unnecessarily hard because the handoff only happens in
ceph_osdc_start_request() instead of when the pagelist is passed to
ceph_osd_data_pagelist_init().  I counted several memory leaks on
various error paths.

Fix up ceph_msg_data_add_pagelist() and carry a pagelist ref in
ceph_osd_data.

Signed-off-by: Ilya Dryomov <idryomov@gmail.com>

diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c
index 580a79b9a91fd9c74dfbe80320b692e089c02c8d..97de674ea3779af49425e6794c69af3b6f30e67b 100644 (file)
--- a/fs/ceph/mds_client.c
+++ b/fs/ceph/mds_client.c
@@ -2136,7 +2136,6 @@ static struct ceph_msg *create_request_message(struct ceph_mds_client *mdsc,
 
        if (req->r_pagelist) {
                struct ceph_pagelist *pagelist = req->r_pagelist;
-               refcount_inc(&pagelist->refcnt);
                ceph_msg_data_add_pagelist(msg, pagelist);
                msg->hdr.data_len = cpu_to_le32(pagelist->length);
        } else {
@@ -3240,6 +3239,7 @@ static void send_mds_reconnect(struct ceph_mds_client *mdsc,
        mutex_unlock(&mdsc->mutex);
 
        up_read(&mdsc->snap_rwsem);
+       ceph_pagelist_release(pagelist);
        return;
 
 fail:

diff --git a/net/ceph/messenger.c b/net/ceph/messenger.c
index 0a187196aeede6375b817a102890f024b552bc3f..76684edc43efd5341f27a7a583171ce26c81106f 100644 (file)
--- a/net/ceph/messenger.c
+++ b/net/ceph/messenger.c
@@ -3313,6 +3313,7 @@ void ceph_msg_data_add_pagelist(struct ceph_msg *msg,
 
        data = ceph_msg_data_create(CEPH_MSG_DATA_PAGELIST);
        BUG_ON(!data);
+       refcount_inc(&pagelist->refcnt);
        data->pagelist = pagelist;
 
        list_add_tail(&data->links, &msg->data);

diff --git a/net/ceph/osd_client.c b/net/ceph/osd_client.c
index db2ebc9e5f1f7b96d0c0ff46fe0d1779e74c5f95..ad00495dbd39249c9ea1ce7f333cffdc4ee6aaae 100644 (file)
--- a/net/ceph/osd_client.c
+++ b/net/ceph/osd_client.c
@@ -126,6 +126,9 @@ static void ceph_osd_data_init(struct ceph_osd_data *osd_data)
        osd_data->type = CEPH_OSD_DATA_TYPE_NONE;
 }
 
+/*
+ * Consumes @pages if @own_pages is true.
+ */
 static void ceph_osd_data_pages_init(struct ceph_osd_data *osd_data,
                        struct page **pages, u64 length, u32 alignment,
                        bool pages_from_pool, bool own_pages)
@@ -138,6 +141,9 @@ static void ceph_osd_data_pages_init(struct ceph_osd_data *osd_data,
        osd_data->own_pages = own_pages;
 }
 
+/*
+ * Consumes a ref on @pagelist.
+ */
 static void ceph_osd_data_pagelist_init(struct ceph_osd_data *osd_data,
                        struct ceph_pagelist *pagelist)
 {
@@ -362,6 +368,8 @@ static void ceph_osd_data_release(struct ceph_osd_data *osd_data)
                num_pages = calc_pages_for((u64)osd_data->alignment,
                                                (u64)osd_data->length);
                ceph_release_page_vector(osd_data->pages, num_pages);
+       } else if (osd_data->type == CEPH_OSD_DATA_TYPE_PAGELIST) {
+               ceph_pagelist_release(osd_data->pagelist);
        }
        ceph_osd_data_init(osd_data);
 }