MINOR: queue: remove the px/srv fields from pendconn

Now we directly use p->queue to get to the queue, which is much more
straightforward. The performance on 100 servers and 16 threads
increased from 560k to 574k RPS, or 2.5%.

A lot more simplifications are possible, but the minimum was done at
this point.
This commit is contained in:
Willy Tarreau 2021-06-23 16:43:45 +02:00
parent 8429097c61
commit 51c63f0f0a
2 changed files with 23 additions and 30 deletions

View File

@ -35,8 +35,6 @@ struct pendconn {
unsigned int queue_idx; /* value of proxy/server queue_idx at time of enqueue */
struct stream *strm;
struct queue *queue; /* the queue the entry is queued into */
struct proxy *px;
struct server *srv; /* the server we are waiting for, may be NULL if don't care */
struct server *target; /* the server that was assigned, = srv except if srv==NULL */
struct eb32_node node;
};

View File

@ -133,7 +133,7 @@ unsigned int srv_dynamic_maxconn(const struct server *s)
*/
static void __pendconn_unlink_srv(struct pendconn *p)
{
p->strm->logs.srv_queue_pos += _HA_ATOMIC_LOAD(&p->srv->queue.idx) - p->queue_idx;
p->strm->logs.srv_queue_pos += _HA_ATOMIC_LOAD(&p->queue->idx) - p->queue_idx;
eb32_delete(&p->node);
}
@ -146,7 +146,7 @@ static void __pendconn_unlink_srv(struct pendconn *p)
*/
static void __pendconn_unlink_prx(struct pendconn *p)
{
p->strm->logs.prx_queue_pos += _HA_ATOMIC_LOAD(&p->px->queue.idx) - p->queue_idx;
p->strm->logs.prx_queue_pos += _HA_ATOMIC_LOAD(&p->queue->idx) - p->queue_idx;
eb32_delete(&p->node);
}
@ -156,10 +156,7 @@ static void __pendconn_unlink_prx(struct pendconn *p)
*/
static inline void pendconn_queue_lock(struct pendconn *p)
{
if (p->srv)
HA_SPIN_LOCK(QUEUE_LOCK, &p->srv->queue.lock);
else
HA_SPIN_LOCK(QUEUE_LOCK, &p->px->queue.lock);
HA_SPIN_LOCK(QUEUE_LOCK, &p->queue->lock);
}
/* Unlocks the queue the pendconn element belongs to. This relies on both p->px
@ -168,10 +165,7 @@ static inline void pendconn_queue_lock(struct pendconn *p)
*/
static inline void pendconn_queue_unlock(struct pendconn *p)
{
if (p->srv)
HA_SPIN_UNLOCK(QUEUE_LOCK, &p->srv->queue.lock);
else
HA_SPIN_UNLOCK(QUEUE_LOCK, &p->px->queue.lock);
HA_SPIN_UNLOCK(QUEUE_LOCK, &p->queue->lock);
}
/* Removes the pendconn from the server/proxy queue. At this stage, the
@ -184,31 +178,34 @@ static inline void pendconn_queue_unlock(struct pendconn *p)
void pendconn_unlink(struct pendconn *p)
{
int done = 0;
struct queue *q = p->queue;
struct proxy *px = q->px;
struct server *sv = q->sv;
if (p->srv) {
if (sv) {
/* queued in the server */
HA_SPIN_LOCK(QUEUE_LOCK, &p->srv->queue.lock);
HA_SPIN_LOCK(QUEUE_LOCK, &q->lock);
if (p->node.node.leaf_p) {
__pendconn_unlink_srv(p);
done = 1;
}
HA_SPIN_UNLOCK(QUEUE_LOCK, &p->srv->queue.lock);
HA_SPIN_UNLOCK(QUEUE_LOCK, &q->lock);
if (done) {
_HA_ATOMIC_DEC(&p->srv->queue.length);
_HA_ATOMIC_DEC(&p->px->totpend);
_HA_ATOMIC_DEC(&q->length);
_HA_ATOMIC_DEC(&px->totpend);
}
}
else {
/* queued in the proxy */
HA_SPIN_LOCK(QUEUE_LOCK, &p->px->queue.lock);
HA_SPIN_LOCK(QUEUE_LOCK, &q->lock);
if (p->node.node.leaf_p) {
__pendconn_unlink_prx(p);
done = 1;
}
HA_SPIN_UNLOCK(QUEUE_LOCK, &p->px->queue.lock);
HA_SPIN_UNLOCK(QUEUE_LOCK, &q->lock);
if (done) {
_HA_ATOMIC_DEC(&p->px->queue.length);
_HA_ATOMIC_DEC(&p->px->totpend);
_HA_ATOMIC_DEC(&q->length);
_HA_ATOMIC_DEC(&px->totpend);
}
}
}
@ -411,20 +408,18 @@ struct pendconn *pendconn_add(struct stream *strm)
if (!p)
return NULL;
p->target = NULL;
p->node.key = MAKE_KEY(strm->priority_class, strm->priority_offset);
p->strm = strm;
p->strm_flags = strm->flags;
strm->pend_pos = p;
px = strm->be;
if (strm->flags & SF_ASSIGNED)
srv = objt_server(strm->target);
else
srv = NULL;
px = strm->be;
p->target = NULL;
p->srv = srv;
p->node.key = MAKE_KEY(strm->priority_class, strm->priority_offset);
p->px = px;
p->strm = strm;
p->strm_flags = strm->flags;
strm->pend_pos = p;
if (srv) {
q = &srv->queue;
max_ptr = &srv->counters.nbpend_max;