summaryrefslogtreecommitdiffstats
path: root/net/ceph/osd_client.c
diff options
context:
space:
mode:
Diffstat (limited to 'net/ceph/osd_client.c')
-rw-r--r--net/ceph/osd_client.c60
1 files changed, 38 insertions, 22 deletions
diff --git a/net/ceph/osd_client.c b/net/ceph/osd_client.c
index ce310eee708d..733e46008b89 100644
--- a/net/ceph/osd_client.c
+++ b/net/ceph/osd_client.c
@@ -217,6 +217,7 @@ struct ceph_osd_request *ceph_osdc_alloc_request(struct ceph_osd_client *osdc,
INIT_LIST_HEAD(&req->r_unsafe_item);
INIT_LIST_HEAD(&req->r_linger_item);
INIT_LIST_HEAD(&req->r_linger_osd);
+ INIT_LIST_HEAD(&req->r_req_lru_item);
req->r_flags = flags;
WARN_ON((flags & (CEPH_OSD_FLAG_READ|CEPH_OSD_FLAG_WRITE)) == 0);
@@ -226,7 +227,7 @@ struct ceph_osd_request *ceph_osdc_alloc_request(struct ceph_osd_client *osdc,
msg = ceph_msgpool_get(&osdc->msgpool_op_reply, 0);
else
msg = ceph_msg_new(CEPH_MSG_OSD_OPREPLY,
- OSD_OPREPLY_FRONT_LEN, gfp_flags);
+ OSD_OPREPLY_FRONT_LEN, gfp_flags, true);
if (!msg) {
ceph_osdc_put_request(req);
return NULL;
@@ -249,7 +250,7 @@ struct ceph_osd_request *ceph_osdc_alloc_request(struct ceph_osd_client *osdc,
if (use_mempool)
msg = ceph_msgpool_get(&osdc->msgpool_op, 0);
else
- msg = ceph_msg_new(CEPH_MSG_OSD_OP, msg_size, gfp_flags);
+ msg = ceph_msg_new(CEPH_MSG_OSD_OP, msg_size, gfp_flags, true);
if (!msg) {
ceph_osdc_put_request(req);
return NULL;
@@ -685,6 +686,18 @@ static void __remove_osd(struct ceph_osd_client *osdc, struct ceph_osd *osd)
put_osd(osd);
}
+static void remove_all_osds(struct ceph_osd_client *osdc)
+{
+ dout("__remove_old_osds %p\n", osdc);
+ mutex_lock(&osdc->request_mutex);
+ while (!RB_EMPTY_ROOT(&osdc->osds)) {
+ struct ceph_osd *osd = rb_entry(rb_first(&osdc->osds),
+ struct ceph_osd, o_node);
+ __remove_osd(osdc, osd);
+ }
+ mutex_unlock(&osdc->request_mutex);
+}
+
static void __move_osd_to_lru(struct ceph_osd_client *osdc,
struct ceph_osd *osd)
{
@@ -701,14 +714,14 @@ static void __remove_osd_from_lru(struct ceph_osd *osd)
list_del_init(&osd->o_osd_lru);
}
-static void remove_old_osds(struct ceph_osd_client *osdc, int remove_all)
+static void remove_old_osds(struct ceph_osd_client *osdc)
{
struct ceph_osd *osd, *nosd;
dout("__remove_old_osds %p\n", osdc);
mutex_lock(&osdc->request_mutex);
list_for_each_entry_safe(osd, nosd, &osdc->osd_lru, o_osd_lru) {
- if (!remove_all && time_before(jiffies, osd->lru_ttl))
+ if (time_before(jiffies, osd->lru_ttl))
break;
__remove_osd(osdc, osd);
}
@@ -751,6 +764,7 @@ static void __insert_osd(struct ceph_osd_client *osdc, struct ceph_osd *new)
struct rb_node *parent = NULL;
struct ceph_osd *osd = NULL;
+ dout("__insert_osd %p osd%d\n", new, new->o_osd);
while (*p) {
parent = *p;
osd = rb_entry(parent, struct ceph_osd, o_node);
@@ -803,13 +817,10 @@ static void __register_request(struct ceph_osd_client *osdc,
{
req->r_tid = ++osdc->last_tid;
req->r_request->hdr.tid = cpu_to_le64(req->r_tid);
- INIT_LIST_HEAD(&req->r_req_lru_item);
-
dout("__register_request %p tid %lld\n", req, req->r_tid);
__insert_request(osdc, req);
ceph_osdc_get_request(req);
osdc->num_requests++;
-
if (osdc->num_requests == 1) {
dout(" first request, scheduling timeout\n");
__schedule_osd_timeout(osdc);
@@ -932,7 +943,7 @@ EXPORT_SYMBOL(ceph_osdc_set_request_linger);
* Caller should hold map_sem for read and request_mutex.
*/
static int __map_request(struct ceph_osd_client *osdc,
- struct ceph_osd_request *req)
+ struct ceph_osd_request *req, int force_resend)
{
struct ceph_osd_request_head *reqhead = req->r_request->front.iov_base;
struct ceph_pg pgid;
@@ -956,7 +967,8 @@ static int __map_request(struct ceph_osd_client *osdc,
num = err;
}
- if ((req->r_osd && req->r_osd->o_osd == o &&
+ if ((!force_resend &&
+ req->r_osd && req->r_osd->o_osd == o &&
req->r_sent >= req->r_osd->o_incarnation &&
req->r_num_pg_osds == num &&
memcmp(req->r_pg_osds, acting, sizeof(acting[0])*num) == 0) ||
@@ -1144,7 +1156,7 @@ static void handle_osds_timeout(struct work_struct *work)
dout("osds timeout\n");
down_read(&osdc->map_sem);
- remove_old_osds(osdc, 0);
+ remove_old_osds(osdc);
up_read(&osdc->map_sem);
schedule_delayed_work(&osdc->osds_timeout_work,
@@ -1278,18 +1290,18 @@ static void reset_changed_osds(struct ceph_osd_client *osdc)
*
* Caller should hold map_sem for read and request_mutex.
*/
-static void kick_requests(struct ceph_osd_client *osdc)
+static void kick_requests(struct ceph_osd_client *osdc, int force_resend)
{
struct ceph_osd_request *req, *nreq;
struct rb_node *p;
int needmap = 0;
int err;
- dout("kick_requests\n");
+ dout("kick_requests %s\n", force_resend ? " (force resend)" : "");
mutex_lock(&osdc->request_mutex);
for (p = rb_first(&osdc->requests); p; p = rb_next(p)) {
req = rb_entry(p, struct ceph_osd_request, r_node);
- err = __map_request(osdc, req);
+ err = __map_request(osdc, req, force_resend);
if (err < 0)
continue; /* error */
if (req->r_osd == NULL) {
@@ -1307,7 +1319,7 @@ static void kick_requests(struct ceph_osd_client *osdc)
r_linger_item) {
dout("linger req=%p req->r_osd=%p\n", req, req->r_osd);
- err = __map_request(osdc, req);
+ err = __map_request(osdc, req, force_resend);
if (err == 0)
continue; /* no change and no osd was specified */
if (err < 0)
@@ -1384,7 +1396,7 @@ void ceph_osdc_handle_map(struct ceph_osd_client *osdc, struct ceph_msg *msg)
ceph_osdmap_destroy(osdc->osdmap);
osdc->osdmap = newmap;
}
- kick_requests(osdc);
+ kick_requests(osdc, 0);
reset_changed_osds(osdc);
} else {
dout("ignoring incremental map %u len %d\n",
@@ -1412,6 +1424,8 @@ void ceph_osdc_handle_map(struct ceph_osd_client *osdc, struct ceph_msg *msg)
"older than our %u\n", epoch, maplen,
osdc->osdmap->epoch);
} else {
+ int skipped_map = 0;
+
dout("taking full map %u len %d\n", epoch, maplen);
newmap = osdmap_decode(&p, p+maplen);
if (IS_ERR(newmap)) {
@@ -1421,9 +1435,12 @@ void ceph_osdc_handle_map(struct ceph_osd_client *osdc, struct ceph_msg *msg)
BUG_ON(!newmap);
oldmap = osdc->osdmap;
osdc->osdmap = newmap;
- if (oldmap)
+ if (oldmap) {
+ if (oldmap->epoch + 1 < newmap->epoch)
+ skipped_map = 1;
ceph_osdmap_destroy(oldmap);
- kick_requests(osdc);
+ }
+ kick_requests(osdc, skipped_map);
}
p += maplen;
nr_maps--;
@@ -1696,7 +1713,7 @@ int ceph_osdc_start_request(struct ceph_osd_client *osdc,
* the request still han't been touched yet.
*/
if (req->r_sent == 0) {
- rc = __map_request(osdc, req);
+ rc = __map_request(osdc, req, 0);
if (rc < 0) {
if (nofail) {
dout("osdc_start_request failed map, "
@@ -1862,8 +1879,7 @@ void ceph_osdc_stop(struct ceph_osd_client *osdc)
ceph_osdmap_destroy(osdc->osdmap);
osdc->osdmap = NULL;
}
- remove_old_osds(osdc, 1);
- WARN_ON(!RB_EMPTY_ROOT(&osdc->osds));
+ remove_all_osds(osdc);
mempool_destroy(osdc->req_mempool);
ceph_msgpool_destroy(&osdc->msgpool_op);
ceph_msgpool_destroy(&osdc->msgpool_op_reply);
@@ -2022,7 +2038,7 @@ static struct ceph_msg *get_reply(struct ceph_connection *con,
if (front > req->r_reply->front.iov_len) {
pr_warning("get_reply front %d > preallocated %d\n",
front, (int)req->r_reply->front.iov_len);
- m = ceph_msg_new(CEPH_MSG_OSD_OPREPLY, front, GFP_NOFS);
+ m = ceph_msg_new(CEPH_MSG_OSD_OPREPLY, front, GFP_NOFS, false);
if (!m)
goto out;
ceph_msg_put(req->r_reply);
@@ -2070,7 +2086,7 @@ static struct ceph_msg *alloc_msg(struct ceph_connection *con,
switch (type) {
case CEPH_MSG_OSD_MAP:
case CEPH_MSG_WATCH_NOTIFY:
- return ceph_msg_new(type, front, GFP_NOFS);
+ return ceph_msg_new(type, front, GFP_NOFS, false);
case CEPH_MSG_OSD_OPREPLY:
return get_reply(con, hdr, skip);
default:
generated by cgit v1.2.3 (git 2.46.0) at 2025-04-25 16:08:07 +0000