Hi
I tried the CFQ scheduler patch you posted on top of 2.6.0-test4-mm3-1,
including the O19int patch from Con Kolivas and I got the attached stacktrace.
---------------------------------------------
This message was sent using M-Web Airmail - JUST LIKE THAT
M-Web: S.A.'s most trusted and reliable Internet Service Provider.
http://airmail.mweb.co.za/
On Mon, Sep 01 2003, Bongani Hlope wrote:
> Hi
>
> I tried the CFQ scheduler patch you posted on top of 2.6.0-test4-mm3-1,
> including the O19int patch from Con Kolivas and I got the attached stacktrace.
Thanks Bongani, I reproduced this problem yesterday. I'll post an update
later today. There appears to be a missing last_merge clear somewhere,
odd.
--
Jens Axboe
On Tue, Sep 02 2003, Jens Axboe wrote:
> On Mon, Sep 01 2003, Bongani Hlope wrote:
> > Hi
> >
> > I tried the CFQ scheduler patch you posted on top of 2.6.0-test4-mm3-1,
> > including the O19int patch from Con Kolivas and I got the attached stacktrace.
>
> Thanks Bongani, I reproduced this problem yesterday. I'll post an update
> later today. There appears to be a missing last_merge clear somewhere,
> odd.
This should be better, delete the one I sent an hour ago.
===== drivers/block/cfq-iosched.c 1.1 vs edited =====
--- 1.1/drivers/block/cfq-iosched.c Fri Aug 29 10:12:50 2003
+++ edited/drivers/block/cfq-iosched.c Tue Sep 2 15:20:05 2003
@@ -149,15 +149,17 @@
/*
* rb tree support functions
*/
-#define RB_EMPTY(root) ((root)->rb_node == NULL)
-#define RB_CLEAR(root) ((root)->rb_node = NULL)
-#define ON_RB(crq) ((crq)->cfq_queue != NULL)
+#define RB_NONE (2)
+#define RB_EMPTY(node) ((node)->rb_node == NULL)
+#define RB_CLEAR(node) ((node)->rb_color = RB_NONE)
+#define RB_CLEAR_ROOT(root) ((root)->rb_node = NULL)
+#define ON_RB(node) ((node)->rb_color != RB_NONE)
#define rb_entry_crq(node) rb_entry((node), struct cfq_rq, rb_node)
#define rq_rb_key(rq) (rq)->sector
static inline void cfq_del_crq_rb(struct cfq_queue *cfqq, struct cfq_rq *crq)
{
- if (ON_RB(crq)) {
+ if (ON_RB(&crq->rb_node)) {
cfqq->queued[rq_data_dir(crq->request)]--;
rb_erase(&crq->rb_node, &cfqq->sort_list);
crq->cfq_queue = NULL;
@@ -194,13 +196,12 @@
struct cfq_rq *__alias;
crq->rb_key = rq_rb_key(rq);
-
+ cfqq->queued[rq_data_dir(rq)]++;
retry:
__alias = __cfq_add_crq_rb(cfqq, crq);
if (!__alias) {
rb_insert_color(&crq->rb_node, &cfqq->sort_list);
crq->cfq_queue = cfqq;
- cfqq->queued[rq_data_dir(rq)]++;
return;
}
@@ -301,7 +302,7 @@
cfq_del_crq_hash(crq);
cfq_add_crq_hash(cfqd, crq);
- if (ON_RB(crq) && rq_rb_key(req) != crq->rb_key) {
+ if (ON_RB(&crq->rb_node) && (rq_rb_key(req) != crq->rb_key)) {
struct cfq_queue *cfqq = crq->cfq_queue;
cfq_del_crq_rb(cfqq, crq);
@@ -345,17 +346,17 @@
}
static inline void
-__cfq_dispatch_requests(struct cfq_data *cfqd, struct cfq_queue *cfqq)
+__cfq_dispatch_requests(request_queue_t *q, struct cfq_data *cfqd,
+ struct cfq_queue *cfqq)
{
- struct cfq_rq *crq;
-
- crq = rb_entry_crq(rb_first(&cfqq->sort_list));
+ struct cfq_rq *crq = rb_entry_crq(rb_first(&cfqq->sort_list));
cfq_del_crq_rb(cfqq, crq);
+ cfq_remove_merge_hints(q, crq);
cfq_dispatch_sort(cfqd->dispatch, crq);
}
-static int cfq_dispatch_requests(struct cfq_data *cfqd)
+static int cfq_dispatch_requests(request_queue_t *q, struct cfq_data *cfqd)
{
struct cfq_queue *cfqq;
struct list_head *entry, *tmp;
@@ -372,7 +373,7 @@
BUG_ON(RB_EMPTY(&cfqq->sort_list));
- __cfq_dispatch_requests(cfqd, cfqq);
+ __cfq_dispatch_requests(q, cfqd, cfqq);
if (RB_EMPTY(&cfqq->sort_list))
cfq_put_queue(cfqd, cfqq);
@@ -399,17 +400,15 @@
dispatch:
rq = list_entry_rq(cfqd->dispatch->next);
- if (q->last_merge == &rq->queuelist)
- q->last_merge = NULL;
-
+ BUG_ON(q->last_merge == &rq->queuelist);
crq = RQ_DATA(rq);
if (crq)
- cfq_del_crq_hash(crq);
+ BUG_ON(ON_MHASH(crq));
return rq;
}
- if (cfq_dispatch_requests(cfqd))
+ if (cfq_dispatch_requests(q, cfqd))
goto dispatch;
return NULL;
@@ -456,7 +455,7 @@
INIT_LIST_HEAD(&cfqq->cfq_hash);
INIT_LIST_HEAD(&cfqq->cfq_list);
- RB_CLEAR(&cfqq->sort_list);
+ RB_CLEAR_ROOT(&cfqq->sort_list);
cfqq->pid = pid;
cfqq->queued[0] = cfqq->queued[1] = 0;
@@ -552,10 +551,9 @@
if (!cfqq)
goto out;
- if (cfqq->queued[rw] < cfq_queued)
- goto out;
-
limit = (q->nr_requests - cfq_queued) / cfqd->busy_queues;
+ if (limit < 3)
+ limit = 3;
if (cfqq->queued[rw] > limit)
ret = 0;
@@ -583,6 +581,7 @@
struct cfq_rq *crq = mempool_alloc(cfqd->crq_pool, gfp_mask);
if (crq) {
+ RB_CLEAR(&crq->rb_node);
crq->request = rq;
crq->cfq_queue = NULL;
INIT_LIST_HEAD(&crq->hash);
--
Jens Axboe
On Tue, Sep 02 2003, Jens Axboe wrote:
> On Mon, Sep 01 2003, Bongani Hlope wrote:
> > Hi
> >
> > I tried the CFQ scheduler patch you posted on top of 2.6.0-test4-mm3-1,
> > including the O19int patch from Con Kolivas and I got the attached stacktrace.
>
> Thanks Bongani, I reproduced this problem yesterday. I'll post an update
> later today. There appears to be a missing last_merge clear somewhere,
> odd.
Can you try this as an incremental patch? It cleans a few things up.
===== drivers/block/cfq-iosched.c 1.1 vs edited =====
--- 1.1/drivers/block/cfq-iosched.c Fri Aug 29 10:12:50 2003
+++ edited/drivers/block/cfq-iosched.c Tue Sep 2 14:29:58 2003
@@ -149,15 +149,17 @@
/*
* rb tree support functions
*/
-#define RB_EMPTY(root) ((root)->rb_node == NULL)
-#define RB_CLEAR(root) ((root)->rb_node = NULL)
-#define ON_RB(crq) ((crq)->cfq_queue != NULL)
+#define RB_NONE (2)
+#define RB_EMPTY(node) ((node)->rb_node == NULL)
+#define RB_CLEAR(node) ((node)->rb_color = RB_NONE)
+#define RB_CLEAR_ROOT(root) ((root)->rb_node = NULL)
+#define ON_RB(node) ((node)->rb_color != RB_NONE)
#define rb_entry_crq(node) rb_entry((node), struct cfq_rq, rb_node)
#define rq_rb_key(rq) (rq)->sector
static inline void cfq_del_crq_rb(struct cfq_queue *cfqq, struct cfq_rq *crq)
{
- if (ON_RB(crq)) {
+ if (ON_RB(&crq->rb_node)) {
cfqq->queued[rq_data_dir(crq->request)]--;
rb_erase(&crq->rb_node, &cfqq->sort_list);
crq->cfq_queue = NULL;
@@ -301,7 +303,7 @@
cfq_del_crq_hash(crq);
cfq_add_crq_hash(cfqd, crq);
- if (ON_RB(crq) && rq_rb_key(req) != crq->rb_key) {
+ if (ON_RB(&crq->rb_node) && (rq_rb_key(req) != crq->rb_key)) {
struct cfq_queue *cfqq = crq->cfq_queue;
cfq_del_crq_rb(cfqq, crq);
@@ -345,17 +347,17 @@
}
static inline void
-__cfq_dispatch_requests(struct cfq_data *cfqd, struct cfq_queue *cfqq)
+__cfq_dispatch_requests(request_queue_t *q, struct cfq_data *cfqd,
+ struct cfq_queue *cfqq)
{
- struct cfq_rq *crq;
-
- crq = rb_entry_crq(rb_first(&cfqq->sort_list));
+ struct cfq_rq *crq = rb_entry_crq(rb_first(&cfqq->sort_list));
cfq_del_crq_rb(cfqq, crq);
+ cfq_remove_merge_hints(q, crq);
cfq_dispatch_sort(cfqd->dispatch, crq);
}
-static int cfq_dispatch_requests(struct cfq_data *cfqd)
+static int cfq_dispatch_requests(request_queue_t *q, struct cfq_data *cfqd)
{
struct cfq_queue *cfqq;
struct list_head *entry, *tmp;
@@ -372,7 +374,7 @@
BUG_ON(RB_EMPTY(&cfqq->sort_list));
- __cfq_dispatch_requests(cfqd, cfqq);
+ __cfq_dispatch_requests(q, cfqd, cfqq);
if (RB_EMPTY(&cfqq->sort_list))
cfq_put_queue(cfqd, cfqq);
@@ -399,17 +401,15 @@
dispatch:
rq = list_entry_rq(cfqd->dispatch->next);
- if (q->last_merge == &rq->queuelist)
- q->last_merge = NULL;
-
+ BUG_ON(q->last_merge == &rq->queuelist);
crq = RQ_DATA(rq);
if (crq)
- cfq_del_crq_hash(crq);
+ BUG_ON(ON_MHASH(crq));
return rq;
}
- if (cfq_dispatch_requests(cfqd))
+ if (cfq_dispatch_requests(q, cfqd))
goto dispatch;
return NULL;
@@ -456,7 +456,7 @@
INIT_LIST_HEAD(&cfqq->cfq_hash);
INIT_LIST_HEAD(&cfqq->cfq_list);
- RB_CLEAR(&cfqq->sort_list);
+ RB_CLEAR_ROOT(&cfqq->sort_list);
cfqq->pid = pid;
cfqq->queued[0] = cfqq->queued[1] = 0;
@@ -583,6 +583,7 @@
struct cfq_rq *crq = mempool_alloc(cfqd->crq_pool, gfp_mask);
if (crq) {
+ RB_CLEAR(&crq->rb_node);
crq->request = rq;
crq->cfq_queue = NULL;
INIT_LIST_HEAD(&crq->hash);
--
Jens Axboe