blkio: Introduce the notion of cfq groups
Vivek Goyal [Thu, 3 Dec 2009 17:59:38 +0000 (12:59 -0500)]
o This patch introduce the notion of cfq groups. Soon we will can have multiple
  groups of different weights in the system.

o Various service trees (prioclass and workload type trees), will become per
  cfq group. So hierarchy looks as follows.

cfq_groups
   |
workload type
   |
        cfq queue

o When an scheduling decision has to be taken, first we select the cfq group
  then workload with-in the group and then cfq queue with-in the workload
  type.

o This patch just makes various workload service tree per cfq group and
  introduce the function to be able to choose a group for scheduling.

Signed-off-by: Vivek Goyal <vgoyal@redhat.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

block/cfq-iosched.c

index 15b5361..a4d1726 100644 (file)
@@ -132,6 +132,7 @@ struct cfq_queue {
 
        struct cfq_rb_root *service_tree;
        struct cfq_queue *new_cfqq;
+       struct cfq_group *cfqg;
 };
 
 /*
@@ -153,25 +154,30 @@ enum wl_type_t {
        SYNC_WORKLOAD = 2
 };
 
+/* This is per cgroup per device grouping structure */
+struct cfq_group {
+       /*
+        * rr lists of queues with requests, onle rr for each priority class.
+        * Counts are embedded in the cfq_rb_root
+        */
+       struct cfq_rb_root service_trees[2][3];
+       struct cfq_rb_root service_tree_idle;
+};
 
 /*
  * Per block device queue structure
  */
 struct cfq_data {
        struct request_queue *queue;
+       struct cfq_group root_group;
 
        /*
-        * rr lists of queues with requests, onle rr for each priority class.
-        * Counts are embedded in the cfq_rb_root
-        */
-       struct cfq_rb_root service_trees[2][3];
-       struct cfq_rb_root service_tree_idle;
-       /*
         * The priority currently being served
         */
        enum wl_prio_t serving_prio;
        enum wl_type_t serving_type;
        unsigned long workload_expires;
+       struct cfq_group *serving_group;
        bool noidle_tree_requires_idle;
 
        /*
@@ -240,14 +246,15 @@ struct cfq_data {
        unsigned long last_end_sync_rq;
 };
 
-static struct cfq_rb_root *service_tree_for(enum wl_prio_t prio,
+static struct cfq_rb_root *service_tree_for(struct cfq_group *cfqg,
+                                           enum wl_prio_t prio,
                                            enum wl_type_t type,
                                            struct cfq_data *cfqd)
 {
        if (prio == IDLE_WORKLOAD)
-               return &cfqd->service_tree_idle;
+               return &cfqg->service_tree_idle;
 
-       return &cfqd->service_trees[prio][type];
+       return &cfqg->service_trees[prio][type];
 }
 
 enum cfqq_state_flags {
@@ -317,12 +324,14 @@ static enum wl_type_t cfqq_type(struct cfq_queue *cfqq)
 
 static inline int cfq_busy_queues_wl(enum wl_prio_t wl, struct cfq_data *cfqd)
 {
+       struct cfq_group *cfqg = &cfqd->root_group;
+
        if (wl == IDLE_WORKLOAD)
-               return cfqd->service_tree_idle.count;
+               return cfqg->service_tree_idle.count;
 
-       return cfqd->service_trees[wl][ASYNC_WORKLOAD].count
-               + cfqd->service_trees[wl][SYNC_NOIDLE_WORKLOAD].count
-               + cfqd->service_trees[wl][SYNC_WORKLOAD].count;
+       return cfqg->service_trees[wl][ASYNC_WORKLOAD].count
+               + cfqg->service_trees[wl][SYNC_NOIDLE_WORKLOAD].count
+               + cfqg->service_trees[wl][SYNC_WORKLOAD].count;
 }
 
 static void cfq_dispatch_insert(struct request_queue *, struct request *);
@@ -612,7 +621,7 @@ static unsigned long cfq_slice_offset(struct cfq_data *cfqd,
        /*
         * just an approximation, should be ok.
         */
-       return (cfqd->busy_queues - 1) * (cfq_prio_slice(cfqd, 1, 0) -
+       return (cfqq->cfqg->nr_cfqq - 1) * (cfq_prio_slice(cfqd, 1, 0) -
                       cfq_prio_slice(cfqd, cfq_cfqq_sync(cfqq), cfqq->ioprio));
 }
 
@@ -630,7 +639,8 @@ static void cfq_service_tree_add(struct cfq_data *cfqd, struct cfq_queue *cfqq,
        struct cfq_rb_root *service_tree;
        int left;
 
-       service_tree = service_tree_for(cfqq_prio(cfqq), cfqq_type(cfqq), cfqd);
+       service_tree = service_tree_for(cfqq->cfqg, cfqq_prio(cfqq),
+                                               cfqq_type(cfqq), cfqd);
        if (cfq_class_idle(cfqq)) {
                rb_key = CFQ_IDLE_DELAY;
                parent = rb_last(&service_tree->rb);
@@ -1066,7 +1076,8 @@ static inline void cfq_slice_expired(struct cfq_data *cfqd, bool timed_out)
 static struct cfq_queue *cfq_get_next_queue(struct cfq_data *cfqd)
 {
        struct cfq_rb_root *service_tree =
-               service_tree_for(cfqd->serving_prio, cfqd->serving_type, cfqd);
+               service_tree_for(cfqd->serving_group, cfqd->serving_prio,
+                                       cfqd->serving_type, cfqd);
 
        if (RB_EMPTY_ROOT(&service_tree->rb))
                return NULL;
@@ -1218,7 +1229,8 @@ static bool cfq_should_idle(struct cfq_data *cfqd, struct cfq_queue *cfqq)
         * in their service tree.
         */
        if (!service_tree)
-               service_tree = service_tree_for(prio, cfqq_type(cfqq), cfqd);
+               service_tree = service_tree_for(cfqq->cfqg, prio,
+                                               cfqq_type(cfqq), cfqd);
 
        if (service_tree->count == 0)
                return true;
@@ -1377,8 +1389,9 @@ static void cfq_setup_merge(struct cfq_queue *cfqq, struct cfq_queue *new_cfqq)
        }
 }
 
-static enum wl_type_t cfq_choose_wl(struct cfq_data *cfqd, enum wl_prio_t prio,
-                                   bool prio_changed)
+static enum wl_type_t cfq_choose_wl(struct cfq_data *cfqd,
+                               struct cfq_group *cfqg, enum wl_prio_t prio,
+                               bool prio_changed)
 {
        struct cfq_queue *queue;
        int i;
@@ -1392,10 +1405,10 @@ static enum wl_type_t cfq_choose_wl(struct cfq_data *cfqd, enum wl_prio_t prio,
                 * from SYNC_NOIDLE (first choice), or just SYNC
                 * over ASYNC
                 */
-               if (service_tree_for(prio, cur_best, cfqd)->count)
+               if (service_tree_for(cfqg, prio, cur_best, cfqd)->count)
                        return cur_best;
                cur_best = SYNC_WORKLOAD;
-               if (service_tree_for(prio, cur_best, cfqd)->count)
+               if (service_tree_for(cfqg, prio, cur_best, cfqd)->count)
                        return cur_best;
 
                return ASYNC_WORKLOAD;
@@ -1403,7 +1416,7 @@ static enum wl_type_t cfq_choose_wl(struct cfq_data *cfqd, enum wl_prio_t prio,
 
        for (i = 0; i < 3; ++i) {
                /* otherwise, select the one with lowest rb_key */
-               queue = cfq_rb_first(service_tree_for(prio, i, cfqd));
+               queue = cfq_rb_first(service_tree_for(cfqg, prio, i, cfqd));
                if (queue &&
                    (!key_valid || time_before(queue->rb_key, lowest_key))) {
                        lowest_key = queue->rb_key;
@@ -1415,12 +1428,13 @@ static enum wl_type_t cfq_choose_wl(struct cfq_data *cfqd, enum wl_prio_t prio,
        return cur_best;
 }
 
-static void choose_service_tree(struct cfq_data *cfqd)
+static void choose_service_tree(struct cfq_data *cfqd, struct cfq_group *cfqg)
 {
        enum wl_prio_t previous_prio = cfqd->serving_prio;
        bool prio_changed;
        unsigned slice;
        unsigned count;
+       struct cfq_rb_root *st;
 
        /* Choose next priority. RT > BE > IDLE */
        if (cfq_busy_queues_wl(RT_WORKLOAD, cfqd))
@@ -1439,8 +1453,9 @@ static void choose_service_tree(struct cfq_data *cfqd)
         * expiration time
         */
        prio_changed = (cfqd->serving_prio != previous_prio);
-       count = service_tree_for(cfqd->serving_prio, cfqd->serving_type, cfqd)
-               ->count;
+       st = service_tree_for(cfqg, cfqd->serving_prio, cfqd->serving_type,
+                               cfqd);
+       count = st->count;
 
        /*
         * If priority didn't change, check workload expiration,
@@ -1452,9 +1467,10 @@ static void choose_service_tree(struct cfq_data *cfqd)
 
        /* otherwise select new workload type */
        cfqd->serving_type =
-               cfq_choose_wl(cfqd, cfqd->serving_prio, prio_changed);
-       count = service_tree_for(cfqd->serving_prio, cfqd->serving_type, cfqd)
-               ->count;
+               cfq_choose_wl(cfqd, cfqg, cfqd->serving_prio, prio_changed);
+       st = service_tree_for(cfqg, cfqd->serving_prio, cfqd->serving_type,
+                               cfqd);
+       count = st->count;
 
        /*
         * the workload slice is computed as a fraction of target latency
@@ -1478,6 +1494,12 @@ static void choose_service_tree(struct cfq_data *cfqd)
        cfqd->noidle_tree_requires_idle = false;
 }
 
+static void cfq_choose_cfqg(struct cfq_data *cfqd)
+{
+       cfqd->serving_group = &cfqd->root_group;
+       choose_service_tree(cfqd, &cfqd->root_group);
+}
+
 /*
  * Select a queue for service. If we have a current active queue,
  * check whether to continue servicing it, or retrieve and set a new one.
@@ -1535,7 +1557,7 @@ new_queue:
         * service tree
         */
        if (!new_cfqq)
-               choose_service_tree(cfqd);
+               cfq_choose_cfqg(cfqd);
 
        cfqq = cfq_set_active_queue(cfqd, new_cfqq);
 keep_queue:
@@ -1564,13 +1586,15 @@ static int cfq_forced_dispatch(struct cfq_data *cfqd)
        struct cfq_queue *cfqq;
        int dispatched = 0;
        int i, j;
+       struct cfq_group *cfqg = &cfqd->root_group;
+
        for (i = 0; i < 2; ++i)
                for (j = 0; j < 3; ++j)
-                       while ((cfqq = cfq_rb_first(&cfqd->service_trees[i][j]))
+                       while ((cfqq = cfq_rb_first(&cfqg->service_trees[i][j]))
                                != NULL)
                                dispatched += __cfq_forced_dispatch_cfqq(cfqq);
 
-       while ((cfqq = cfq_rb_first(&cfqd->service_tree_idle)) != NULL)
+       while ((cfqq = cfq_rb_first(&cfqg->service_tree_idle)) != NULL)
                dispatched += __cfq_forced_dispatch_cfqq(cfqq);
 
        cfq_slice_expired(cfqd, 0);
@@ -2041,14 +2065,26 @@ static void cfq_init_cfqq(struct cfq_data *cfqd, struct cfq_queue *cfqq,
        cfqq->pid = pid;
 }
 
+static void cfq_link_cfqq_cfqg(struct cfq_queue *cfqq, struct cfq_group *cfqg)
+{
+       cfqq->cfqg = cfqg;
+}
+
+static struct cfq_group *cfq_get_cfqg(struct cfq_data *cfqd, int create)
+{
+       return &cfqd->root_group;
+}
+
 static struct cfq_queue *
 cfq_find_alloc_queue(struct cfq_data *cfqd, bool is_sync,
                     struct io_context *ioc, gfp_t gfp_mask)
 {
        struct cfq_queue *cfqq, *new_cfqq = NULL;
        struct cfq_io_context *cic;
+       struct cfq_group *cfqg;
 
 retry:
+       cfqg = cfq_get_cfqg(cfqd, 1);
        cic = cfq_cic_lookup(cfqd, ioc);
        /* cic always exists here */
        cfqq = cic_to_cfqq(cic, is_sync);
@@ -2079,6 +2115,7 @@ retry:
                if (cfqq) {
                        cfq_init_cfqq(cfqd, cfqq, current->pid, is_sync);
                        cfq_init_prio_data(cfqq, ioc);
+                       cfq_link_cfqq_cfqg(cfqq, cfqg);
                        cfq_log_cfqq(cfqd, cfqq, "alloced");
                } else
                        cfqq = &cfqd->oom_cfqq;
@@ -2931,15 +2968,19 @@ static void *cfq_init_queue(struct request_queue *q)
 {
        struct cfq_data *cfqd;
        int i, j;
+       struct cfq_group *cfqg;
 
        cfqd = kmalloc_node(sizeof(*cfqd), GFP_KERNEL | __GFP_ZERO, q->node);
        if (!cfqd)
                return NULL;
 
+       /* Init root group */
+       cfqg = &cfqd->root_group;
+
        for (i = 0; i < 2; ++i)
                for (j = 0; j < 3; ++j)
-                       cfqd->service_trees[i][j] = CFQ_RB_ROOT;
-       cfqd->service_tree_idle = CFQ_RB_ROOT;
+                       cfqg->service_trees[i][j] = CFQ_RB_ROOT;
+       cfqg->service_tree_idle = CFQ_RB_ROOT;
 
        /*
         * Not strictly needed (since RB_ROOT just clears the node and we
@@ -2956,6 +2997,7 @@ static void *cfq_init_queue(struct request_queue *q)
         */
        cfq_init_cfqq(cfqd, &cfqd->oom_cfqq, 1, 0);
        atomic_inc(&cfqd->oom_cfqq.ref);
+       cfq_link_cfqq_cfqg(&cfqd->oom_cfqq, &cfqd->root_group);
 
        INIT_LIST_HEAD(&cfqd->cic_list);