/*
- * linux/drivers/block/elevator.c
- *
* Block device elevator/IO-scheduler.
*
* Copyright (C) 2000 Andrea Arcangeli <andrea@suse.de> SuSE
#include <linux/init.h>
#include <linux/compiler.h>
#include <linux/delay.h>
+#include <linux/blktrace_api.h>
#include <asm/uaccess.h>
}
EXPORT_SYMBOL(elv_rq_merge_ok);
-inline int elv_try_merge(struct request *__rq, struct bio *bio)
+static inline int elv_try_merge(struct request *__rq, struct bio *bio)
{
int ret = ELEVATOR_NO_MERGE;
return ret;
}
-EXPORT_SYMBOL(elv_try_merge);
static struct elevator_type *elevator_find(const char *name)
{
return e;
}
-static int elevator_attach(request_queue_t *q, struct elevator_type *e,
- struct elevator_queue *eq)
+static void *elevator_init_queue(request_queue_t *q, struct elevator_queue *eq)
{
- int ret = 0;
-
- memset(eq, 0, sizeof(*eq));
- eq->ops = &e->ops;
- eq->elevator_type = e;
+ return eq->ops->elevator_init_fn(q, eq);
+}
+static void elevator_attach(request_queue_t *q, struct elevator_queue *eq,
+ void *data)
+{
q->elevator = eq;
-
- if (eq->ops->elevator_init_fn)
- ret = eq->ops->elevator_init_fn(q, eq);
-
- return ret;
+ eq->elevator_data = data;
}
static char chosen_elevator[16];
-static void elevator_setup_default(void)
+static int __init elevator_setup(char *str)
{
- struct elevator_type *e;
-
/*
- * If default has not been set, use the compiled-in selection.
+ * Be backwards-compatible with previous kernels, so users
+ * won't get the wrong elevator.
*/
- if (!chosen_elevator[0])
- strcpy(chosen_elevator, CONFIG_DEFAULT_IOSCHED);
-
- /*
- * If the given scheduler is not available, fall back to no-op.
- */
- if ((e = elevator_find(chosen_elevator)))
- elevator_put(e);
+ if (!strcmp(str, "as"))
+ strcpy(chosen_elevator, "anticipatory");
else
- strcpy(chosen_elevator, "noop");
+ strncpy(chosen_elevator, str, sizeof(chosen_elevator) - 1);
+ return 1;
}
-static int __init elevator_setup(char *str)
+__setup("elevator=", elevator_setup);
+
+static struct kobj_type elv_ktype;
+
+static elevator_t *elevator_alloc(struct elevator_type *e)
{
- strncpy(chosen_elevator, str, sizeof(chosen_elevator) - 1);
- return 0;
+ elevator_t *eq = kmalloc(sizeof(elevator_t), GFP_KERNEL);
+ if (eq) {
+ memset(eq, 0, sizeof(*eq));
+ eq->ops = &e->ops;
+ eq->elevator_type = e;
+ kobject_init(&eq->kobj);
+ snprintf(eq->kobj.name, KOBJ_NAME_LEN, "%s", "iosched");
+ eq->kobj.ktype = &elv_ktype;
+ mutex_init(&eq->sysfs_lock);
+ } else {
+ elevator_put(e);
+ }
+ return eq;
}
-__setup("elevator=", elevator_setup);
+static void elevator_release(struct kobject *kobj)
+{
+ elevator_t *e = container_of(kobj, elevator_t, kobj);
+ elevator_put(e->elevator_type);
+ kfree(e);
+}
int elevator_init(request_queue_t *q, char *name)
{
struct elevator_type *e = NULL;
struct elevator_queue *eq;
int ret = 0;
+ void *data;
INIT_LIST_HEAD(&q->queue_head);
q->last_merge = NULL;
q->end_sector = 0;
q->boundary_rq = NULL;
- elevator_setup_default();
+ if (name && !(e = elevator_get(name)))
+ return -EINVAL;
- if (!name)
- name = chosen_elevator;
+ if (!e && *chosen_elevator && !(e = elevator_get(chosen_elevator)))
+ printk("I/O scheduler %s not found\n", chosen_elevator);
- e = elevator_get(name);
- if (!e)
- return -EINVAL;
+ if (!e && !(e = elevator_get(CONFIG_DEFAULT_IOSCHED))) {
+ printk("Default I/O scheduler not found, using no-op\n");
+ e = elevator_get("noop");
+ }
- eq = kmalloc(sizeof(struct elevator_queue), GFP_KERNEL);
- if (!eq) {
- elevator_put(e);
+ eq = elevator_alloc(e);
+ if (!eq)
return -ENOMEM;
- }
- ret = elevator_attach(q, e, eq);
- if (ret) {
- kfree(eq);
- elevator_put(e);
+ data = elevator_init_queue(q, eq);
+ if (!data) {
+ kobject_put(&eq->kobj);
+ return -ENOMEM;
}
+ elevator_attach(q, eq, data);
return ret;
}
void elevator_exit(elevator_t *e)
{
+ mutex_lock(&e->sysfs_lock);
if (e->ops->elevator_exit_fn)
e->ops->elevator_exit_fn(e);
+ e->ops = NULL;
+ mutex_unlock(&e->sysfs_lock);
- elevator_put(e->elevator_type);
- e->elevator_type = NULL;
- kfree(e);
+ kobject_put(&e->kobj);
}
/*
rq->flags &= ~REQ_STARTED;
- /*
- * if this is the flush, requeue the original instead and drop the flush
- */
- if (rq->flags & REQ_BAR_FLUSH) {
- clear_bit(QUEUE_FLAG_FLUSH, &q->queue_flags);
- rq = rq->end_io_data;
- }
-
- __elv_add_request(q, rq, ELEVATOR_INSERT_FRONT, 0);
+ elv_insert(q, rq, ELEVATOR_INSERT_REQUEUE);
}
static void elv_drain_elevator(request_queue_t *q)
}
}
-void __elv_add_request(request_queue_t *q, struct request *rq, int where,
- int plug)
+void elv_insert(request_queue_t *q, struct request *rq, int where)
{
- if (rq->flags & (REQ_SOFTBARRIER | REQ_HARDBARRIER)) {
- /*
- * barriers implicitly indicate back insertion
- */
- if (where == ELEVATOR_INSERT_SORT)
- where = ELEVATOR_INSERT_BACK;
-
- /*
- * this request is scheduling boundary, update end_sector
- */
- if (blk_fs_request(rq)) {
- q->end_sector = rq_end_sector(rq);
- q->boundary_rq = rq;
- }
- } else if (!(rq->flags & REQ_ELVPRIV) && where == ELEVATOR_INSERT_SORT)
- where = ELEVATOR_INSERT_BACK;
+ struct list_head *pos;
+ unsigned ordseq;
+ int unplug_it = 1;
- if (plug)
- blk_plug_device(q);
+ blk_add_trace_rq(q, rq, BLK_TA_INSERT);
rq->q = q;
q->elevator->ops->elevator_add_req_fn(q, rq);
break;
+ case ELEVATOR_INSERT_REQUEUE:
+ /*
+ * If ordered flush isn't in progress, we do front
+ * insertion; otherwise, requests should be requeued
+ * in ordseq order.
+ */
+ rq->flags |= REQ_SOFTBARRIER;
+
+ if (q->ordseq == 0) {
+ list_add(&rq->queuelist, &q->queue_head);
+ break;
+ }
+
+ ordseq = blk_ordered_req_seq(rq);
+
+ list_for_each(pos, &q->queue_head) {
+ struct request *pos_rq = list_entry_rq(pos);
+ if (ordseq <= blk_ordered_req_seq(pos_rq))
+ break;
+ }
+
+ list_add_tail(&rq->queuelist, pos);
+ /*
+ * most requeues happen because of a busy condition, don't
+ * force unplug of the queue for that case.
+ */
+ unplug_it = 0;
+ break;
+
default:
printk(KERN_ERR "%s: bad insertion point %d\n",
__FUNCTION__, where);
BUG();
}
- if (blk_queue_plugged(q)) {
+ if (unplug_it && blk_queue_plugged(q)) {
int nrq = q->rq.count[READ] + q->rq.count[WRITE]
- q->in_flight;
}
}
+void __elv_add_request(request_queue_t *q, struct request *rq, int where,
+ int plug)
+{
+ if (q->ordcolor)
+ rq->flags |= REQ_ORDERED_COLOR;
+
+ if (rq->flags & (REQ_SOFTBARRIER | REQ_HARDBARRIER)) {
+ /*
+ * toggle ordered color
+ */
+ if (blk_barrier_rq(rq))
+ q->ordcolor ^= 1;
+
+ /*
+ * barriers implicitly indicate back insertion
+ */
+ if (where == ELEVATOR_INSERT_SORT)
+ where = ELEVATOR_INSERT_BACK;
+
+ /*
+ * this request is scheduling boundary, update
+ * end_sector
+ */
+ if (blk_fs_request(rq)) {
+ q->end_sector = rq_end_sector(rq);
+ q->boundary_rq = rq;
+ }
+ } else if (!(rq->flags & REQ_ELVPRIV) && where == ELEVATOR_INSERT_SORT)
+ where = ELEVATOR_INSERT_BACK;
+
+ if (plug)
+ blk_plug_device(q);
+
+ elv_insert(q, rq, where);
+}
+
void elv_add_request(request_queue_t *q, struct request *rq, int where,
int plug)
{
{
struct request *rq;
- if (unlikely(list_empty(&q->queue_head) &&
- !q->elevator->ops->elevator_dispatch_fn(q, 0)))
- return NULL;
-
- rq = list_entry_rq(q->queue_head.next);
-
- /*
- * if this is a barrier write and the device has to issue a
- * flush sequence to support it, check how far we are
- */
- if (blk_fs_request(rq) && blk_barrier_rq(rq)) {
- BUG_ON(q->ordered == QUEUE_ORDERED_NONE);
+ while (1) {
+ while (!list_empty(&q->queue_head)) {
+ rq = list_entry_rq(q->queue_head.next);
+ if (blk_do_ordered(q, &rq))
+ return rq;
+ }
- if (q->ordered == QUEUE_ORDERED_FLUSH &&
- !blk_barrier_preflush(rq))
- rq = blk_start_pre_flush(q, rq);
+ if (!q->elevator->ops->elevator_dispatch_fn(q, 0))
+ return NULL;
}
-
- return rq;
}
struct request *elv_next_request(request_queue_t *q)
* not be passed by new incoming requests
*/
rq->flags |= REQ_STARTED;
+ blk_add_trace_rq(q, rq, BLK_TA_ISSUE);
}
if (!q->boundary_rq || q->boundary_rq == rq) {
blkdev_dequeue_request(rq);
rq->flags |= REQ_QUIET;
end_that_request_chunk(rq, 0, nr_bytes);
- end_that_request_last(rq);
+ end_that_request_last(rq, 0);
} else {
printk(KERN_ERR "%s: bad return=%d\n", __FUNCTION__,
ret);
struct request *elv_latter_request(request_queue_t *q, struct request *rq)
{
- struct list_head *next;
-
elevator_t *e = q->elevator;
if (e->ops->elevator_latter_req_fn)
return e->ops->elevator_latter_req_fn(q, rq);
-
- next = rq->queuelist.next;
- if (next != &q->queue_head && next != &rq->queuelist)
- return list_entry_rq(next);
-
return NULL;
}
struct request *elv_former_request(request_queue_t *q, struct request *rq)
{
- struct list_head *prev;
-
elevator_t *e = q->elevator;
if (e->ops->elevator_former_req_fn)
return e->ops->elevator_former_req_fn(q, rq);
-
- prev = rq->queuelist.prev;
- if (prev != &q->queue_head && prev != &rq->queuelist)
- return list_entry_rq(prev);
-
return NULL;
}
if (blk_sorted_rq(rq) && e->ops->elevator_completed_req_fn)
e->ops->elevator_completed_req_fn(q, rq);
}
+
+ /*
+ * Check if the queue is waiting for fs requests to be
+ * drained for flush sequence.
+ */
+ if (unlikely(q->ordseq)) {
+ struct request *first_rq = list_entry_rq(q->queue_head.next);
+ if (q->in_flight == 0 &&
+ blk_ordered_cur_seq(q) == QUEUE_ORDSEQ_DRAIN &&
+ blk_ordered_req_seq(first_rq) > QUEUE_ORDSEQ_DRAIN) {
+ blk_ordered_complete_seq(q, QUEUE_ORDSEQ_DRAIN, 0);
+ q->request_fn(q);
+ }
+ }
}
-int elv_register_queue(struct request_queue *q)
+#define to_elv(atr) container_of((atr), struct elv_fs_entry, attr)
+
+static ssize_t
+elv_attr_show(struct kobject *kobj, struct attribute *attr, char *page)
{
- elevator_t *e = q->elevator;
+ elevator_t *e = container_of(kobj, elevator_t, kobj);
+ struct elv_fs_entry *entry = to_elv(attr);
+ ssize_t error;
+
+ if (!entry->show)
+ return -EIO;
- e->kobj.parent = kobject_get(&q->kobj);
- if (!e->kobj.parent)
- return -EBUSY;
+ mutex_lock(&e->sysfs_lock);
+ error = e->ops ? entry->show(e, page) : -ENOENT;
+ mutex_unlock(&e->sysfs_lock);
+ return error;
+}
+
+static ssize_t
+elv_attr_store(struct kobject *kobj, struct attribute *attr,
+ const char *page, size_t length)
+{
+ elevator_t *e = container_of(kobj, elevator_t, kobj);
+ struct elv_fs_entry *entry = to_elv(attr);
+ ssize_t error;
- snprintf(e->kobj.name, KOBJ_NAME_LEN, "%s", "iosched");
- e->kobj.ktype = e->elevator_type->elevator_ktype;
+ if (!entry->store)
+ return -EIO;
- return kobject_register(&e->kobj);
+ mutex_lock(&e->sysfs_lock);
+ error = e->ops ? entry->store(e, page, length) : -ENOENT;
+ mutex_unlock(&e->sysfs_lock);
+ return error;
}
-void elv_unregister_queue(struct request_queue *q)
+static struct sysfs_ops elv_sysfs_ops = {
+ .show = elv_attr_show,
+ .store = elv_attr_store,
+};
+
+static struct kobj_type elv_ktype = {
+ .sysfs_ops = &elv_sysfs_ops,
+ .release = elevator_release,
+};
+
+int elv_register_queue(struct request_queue *q)
{
- if (q) {
- elevator_t *e = q->elevator;
- kobject_unregister(&e->kobj);
- kobject_put(&q->kobj);
+ elevator_t *e = q->elevator;
+ int error;
+
+ e->kobj.parent = &q->kobj;
+
+ error = kobject_add(&e->kobj);
+ if (!error) {
+ struct elv_fs_entry *attr = e->elevator_type->elevator_attrs;
+ if (attr) {
+ while (attr->attr.name) {
+ if (sysfs_create_file(&e->kobj, &attr->attr))
+ break;
+ attr++;
+ }
+ }
+ kobject_uevent(&e->kobj, KOBJ_ADD);
}
+ return error;
+}
+
+static void __elv_unregister_queue(elevator_t *e)
+{
+ kobject_uevent(&e->kobj, KOBJ_REMOVE);
+ kobject_del(&e->kobj);
+}
+
+void elv_unregister_queue(struct request_queue *q)
+{
+ if (q)
+ __elv_unregister_queue(q->elevator);
}
int elv_register(struct elevator_type *e)
{
spin_lock_irq(&elv_list_lock);
- if (elevator_find(e->elevator_name))
- BUG();
+ BUG_ON(elevator_find(e->elevator_name));
list_add_tail(&e->list, &elv_list);
spin_unlock_irq(&elv_list_lock);
printk(KERN_INFO "io scheduler %s registered", e->elevator_name);
- if (!strcmp(e->elevator_name, chosen_elevator))
- printk(" (default)");
+ if (!strcmp(e->elevator_name, chosen_elevator) ||
+ (!*chosen_elevator &&
+ !strcmp(e->elevator_name, CONFIG_DEFAULT_IOSCHED)))
+ printk(" (default)");
printk("\n");
return 0;
}
/*
* Iterate every thread in the process to remove the io contexts.
*/
- read_lock(&tasklist_lock);
- do_each_thread(g, p) {
- struct io_context *ioc = p->io_context;
- if (ioc && ioc->cic) {
- ioc->cic->exit(ioc->cic);
- ioc->cic->dtor(ioc->cic);
- ioc->cic = NULL;
- }
- if (ioc && ioc->aic) {
- ioc->aic->exit(ioc->aic);
- ioc->aic->dtor(ioc->aic);
- ioc->aic = NULL;
- }
- } while_each_thread(g, p);
- read_unlock(&tasklist_lock);
+ if (e->ops.trim) {
+ read_lock(&tasklist_lock);
+ do_each_thread(g, p) {
+ task_lock(p);
+ e->ops.trim(p->io_context);
+ task_unlock(p);
+ } while_each_thread(g, p);
+ read_unlock(&tasklist_lock);
+ }
spin_lock_irq(&elv_list_lock);
list_del_init(&e->list);
* need for the new one. this way we have a chance of going back to the old
* one, if the new one fails init for some reason.
*/
-static void elevator_switch(request_queue_t *q, struct elevator_type *new_e)
+static int elevator_switch(request_queue_t *q, struct elevator_type *new_e)
{
elevator_t *old_elevator, *e;
+ void *data;
/*
* Allocate new elevator
*/
- e = kmalloc(sizeof(elevator_t), GFP_KERNEL);
+ e = elevator_alloc(new_e);
if (!e)
- goto error;
+ return 0;
+
+ data = elevator_init_queue(q, e);
+ if (!data) {
+ kobject_put(&e->kobj);
+ return 0;
+ }
/*
* Turn on BYPASS and drain all requests w/ elevator private data
elv_drain_elevator(q);
}
- spin_unlock_irq(q->queue_lock);
-
/*
- * unregister old elevator data
+ * Remember old elevator.
*/
- elv_unregister_queue(q);
old_elevator = q->elevator;
/*
* attach and start new elevator
*/
- if (elevator_attach(q, new_e, e))
- goto fail;
+ elevator_attach(q, e, data);
+
+ spin_unlock_irq(q->queue_lock);
+
+ __elv_unregister_queue(old_elevator);
if (elv_register_queue(q))
goto fail_register;
*/
elevator_exit(old_elevator);
clear_bit(QUEUE_FLAG_ELVSWITCH, &q->queue_flags);
- return;
+ return 1;
fail_register:
/*
* one again (along with re-adding the sysfs dir)
*/
elevator_exit(e);
- e = NULL;
-fail:
q->elevator = old_elevator;
elv_register_queue(q);
clear_bit(QUEUE_FLAG_ELVSWITCH, &q->queue_flags);
- kfree(e);
-error:
- elevator_put(new_e);
- printk(KERN_ERR "elevator: switch to %s failed\n",new_e->elevator_name);
+ return 0;
}
ssize_t elv_iosched_store(request_queue_t *q, const char *name, size_t count)
{
char elevator_name[ELV_NAME_MAX];
+ size_t len;
struct elevator_type *e;
- memset(elevator_name, 0, sizeof(elevator_name));
- strncpy(elevator_name, name, sizeof(elevator_name));
+ elevator_name[sizeof(elevator_name) - 1] = '\0';
+ strncpy(elevator_name, name, sizeof(elevator_name) - 1);
+ len = strlen(elevator_name);
- if (elevator_name[strlen(elevator_name) - 1] == '\n')
- elevator_name[strlen(elevator_name) - 1] = '\0';
+ if (len && elevator_name[len - 1] == '\n')
+ elevator_name[len - 1] = '\0';
e = elevator_get(elevator_name);
if (!e) {
return count;
}
- elevator_switch(q, e);
+ if (!elevator_switch(q, e))
+ printk(KERN_ERR "elevator: switch to %s failed\n",elevator_name);
return count;
}
EXPORT_SYMBOL(elv_dispatch_sort);
EXPORT_SYMBOL(elv_add_request);
EXPORT_SYMBOL(__elv_add_request);
-EXPORT_SYMBOL(elv_requeue_request);
EXPORT_SYMBOL(elv_next_request);
EXPORT_SYMBOL(elv_dequeue_request);
EXPORT_SYMBOL(elv_queue_empty);
-EXPORT_SYMBOL(elv_completed_request);
EXPORT_SYMBOL(elevator_exit);
EXPORT_SYMBOL(elevator_init);