]> www.pilppa.org Git - linux-2.6-omap-h63xx.git/blobdiff - block/blk-core.c
block: make core bits checkpatch compliant
[linux-2.6-omap-h63xx.git] / block / blk-core.c
index 55cf293d907dd0901d4cbf5e0d3279936bd2a2ea..4afb39c823396ccec50c3ded2673d22ed0b07cf3 100644 (file)
@@ -3,7 +3,8 @@
  * Copyright (C) 1994,      Karl Keyte: Added support for disk statistics
  * Elevator latency, (C) 2000  Andrea Arcangeli <andrea@suse.de> SuSE
  * Queue request tables / lock, selectable elevator, Jens Axboe <axboe@suse.de>
- * kernel-doc documentation started by NeilBrown <neilb@cse.unsw.edu.au> -  July2000
+ * kernel-doc documentation started by NeilBrown <neilb@cse.unsw.edu.au>
+ *     -  July2000
  * bio rewrite, highmem i/o, etc, Jens Axboe <axboe@suse.de> - may 2001
  */
 
@@ -42,7 +43,7 @@ struct kmem_cache *request_cachep;
 /*
  * For queue allocation
  */
-struct kmem_cache *blk_requestq_cachep = NULL;
+struct kmem_cache *blk_requestq_cachep;
 
 /*
  * Controlling structure to kblockd
@@ -137,7 +138,7 @@ static void req_bio_endio(struct request *rq, struct bio *bio,
                        error = -EIO;
 
                if (unlikely(nbytes > bio->bi_size)) {
-                       printk("%s: want %u bytes done, only %u left\n",
+                       printk(KERN_ERR "%s: want %u bytes done, %u left\n",
                               __FUNCTION__, nbytes, bio->bi_size);
                        nbytes = bio->bi_size;
                }
@@ -161,23 +162,26 @@ void blk_dump_rq_flags(struct request *rq, char *msg)
 {
        int bit;
 
-       printk("%s: dev %s: type=%x, flags=%x\n", msg,
+       printk(KERN_INFO "%s: dev %s: type=%x, flags=%x\n", msg,
                rq->rq_disk ? rq->rq_disk->disk_name : "?", rq->cmd_type,
                rq->cmd_flags);
 
-       printk("\nsector %llu, nr/cnr %lu/%u\n", (unsigned long long)rq->sector,
-                                                      rq->nr_sectors,
-                                                      rq->current_nr_sectors);
-       printk("bio %p, biotail %p, buffer %p, data %p, len %u\n", rq->bio, rq->biotail, rq->buffer, rq->data, rq->data_len);
+       printk(KERN_INFO "  sector %llu, nr/cnr %lu/%u\n",
+                                               (unsigned long long)rq->sector,
+                                               rq->nr_sectors,
+                                               rq->current_nr_sectors);
+       printk(KERN_INFO "  bio %p, biotail %p, buffer %p, data %p, len %u\n",
+                                               rq->bio, rq->biotail,
+                                               rq->buffer, rq->data,
+                                               rq->data_len);
 
        if (blk_pc_request(rq)) {
-               printk("cdb: ");
+               printk(KERN_INFO "  cdb: ");
                for (bit = 0; bit < sizeof(rq->cmd); bit++)
                        printk("%02x ", rq->cmd[bit]);
                printk("\n");
        }
 }
-
 EXPORT_SYMBOL(blk_dump_rq_flags);
 
 /*
@@ -204,7 +208,6 @@ void blk_plug_device(struct request_queue *q)
                blk_add_trace_generic(q, NULL, 0, BLK_TA_PLUG);
        }
 }
-
 EXPORT_SYMBOL(blk_plug_device);
 
 /*
@@ -221,7 +224,6 @@ int blk_remove_plug(struct request_queue *q)
        del_timer(&q->unplug_timer);
        return 1;
 }
-
 EXPORT_SYMBOL(blk_remove_plug);
 
 /*
@@ -328,7 +330,6 @@ void blk_start_queue(struct request_queue *q)
                kblockd_schedule_work(&q->unplug_work);
        }
 }
-
 EXPORT_SYMBOL(blk_start_queue);
 
 /**
@@ -408,7 +409,7 @@ void blk_put_queue(struct request_queue *q)
 }
 EXPORT_SYMBOL(blk_put_queue);
 
-void blk_cleanup_queue(struct request_queue * q)
+void blk_cleanup_queue(struct request_queue *q)
 {
        mutex_lock(&q->sysfs_lock);
        set_bit(QUEUE_FLAG_DEAD, &q->queue_flags);
@@ -419,7 +420,6 @@ void blk_cleanup_queue(struct request_queue * q)
 
        blk_put_queue(q);
 }
-
 EXPORT_SYMBOL(blk_cleanup_queue);
 
 static int blk_init_free_list(struct request_queue *q)
@@ -575,7 +575,6 @@ int blk_get_queue(struct request_queue *q)
 
        return 1;
 }
-
 EXPORT_SYMBOL(blk_get_queue);
 
 static inline void blk_free_request(struct request_queue *q, struct request *rq)
@@ -774,7 +773,7 @@ rq_starved:
         */
        if (ioc_batching(q, ioc))
                ioc->nr_batch_requests--;
-       
+
        rq_init(q, rq);
 
        blk_add_trace_generic(q, bio, rw, BLK_TA_GETRQ);
@@ -888,7 +887,6 @@ void blk_requeue_request(struct request_queue *q, struct request *rq)
 
        elv_requeue_request(q, rq);
 }
-
 EXPORT_SYMBOL(blk_requeue_request);
 
 /**
@@ -939,7 +937,6 @@ void blk_insert_request(struct request_queue *q, struct request *rq,
        blk_start_queueing(q);
        spin_unlock_irqrestore(q->queue_lock, flags);
 }
-
 EXPORT_SYMBOL(blk_insert_request);
 
 /*
@@ -947,7 +944,7 @@ EXPORT_SYMBOL(blk_insert_request);
  * queue lock is held and interrupts disabled, as we muck with the
  * request queue list.
  */
-static inline void add_request(struct request_queue * q, struct request * req)
+static inline void add_request(struct request_queue *q, struct request *req)
 {
        drive_stat_acct(req, 1);
 
@@ -957,7 +954,7 @@ static inline void add_request(struct request_queue * q, struct request * req)
         */
        __elv_add_request(q, req, ELEVATOR_INSERT_SORT, 0);
 }
+
 /*
  * disk_round_stats()  - Round off the performance stats on a struct
  * disk_stats.
@@ -987,7 +984,6 @@ void disk_round_stats(struct gendisk *disk)
        }
        disk->stamp = now;
 }
-
 EXPORT_SYMBOL_GPL(disk_round_stats);
 
 /*
@@ -1017,7 +1013,6 @@ void __blk_put_request(struct request_queue *q, struct request *req)
                freed_request(q, rw, priv);
        }
 }
-
 EXPORT_SYMBOL_GPL(__blk_put_request);
 
 void blk_put_request(struct request *req)
@@ -1035,7 +1030,6 @@ void blk_put_request(struct request *req)
                spin_unlock_irqrestore(q->queue_lock, flags);
        }
 }
-
 EXPORT_SYMBOL(blk_put_request);
 
 void init_request_from_bio(struct request *req, struct bio *bio)
@@ -1096,53 +1090,53 @@ static int __make_request(struct request_queue *q, struct bio *bio)
 
        el_ret = elv_merge(q, &req, bio);
        switch (el_ret) {
-               case ELEVATOR_BACK_MERGE:
-                       BUG_ON(!rq_mergeable(req));
+       case ELEVATOR_BACK_MERGE:
+               BUG_ON(!rq_mergeable(req));
 
-                       if (!ll_back_merge_fn(q, req, bio))
-                               break;
+               if (!ll_back_merge_fn(q, req, bio))
+                       break;
 
-                       blk_add_trace_bio(q, bio, BLK_TA_BACKMERGE);
+               blk_add_trace_bio(q, bio, BLK_TA_BACKMERGE);
 
-                       req->biotail->bi_next = bio;
-                       req->biotail = bio;
-                       req->nr_sectors = req->hard_nr_sectors += nr_sectors;
-                       req->ioprio = ioprio_best(req->ioprio, prio);
-                       drive_stat_acct(req, 0);
-                       if (!attempt_back_merge(q, req))
-                               elv_merged_request(q, req, el_ret);
-                       goto out;
+               req->biotail->bi_next = bio;
+               req->biotail = bio;
+               req->nr_sectors = req->hard_nr_sectors += nr_sectors;
+               req->ioprio = ioprio_best(req->ioprio, prio);
+               drive_stat_acct(req, 0);
+               if (!attempt_back_merge(q, req))
+                       elv_merged_request(q, req, el_ret);
+               goto out;
 
-               case ELEVATOR_FRONT_MERGE:
-                       BUG_ON(!rq_mergeable(req));
+       case ELEVATOR_FRONT_MERGE:
+               BUG_ON(!rq_mergeable(req));
 
-                       if (!ll_front_merge_fn(q, req, bio))
-                               break;
+               if (!ll_front_merge_fn(q, req, bio))
+                       break;
 
-                       blk_add_trace_bio(q, bio, BLK_TA_FRONTMERGE);
+               blk_add_trace_bio(q, bio, BLK_TA_FRONTMERGE);
 
-                       bio->bi_next = req->bio;
-                       req->bio = bio;
+               bio->bi_next = req->bio;
+               req->bio = bio;
 
-                       /*
-                        * may not be valid. if the low level driver said
-                        * it didn't need a bounce buffer then it better
-                        * not touch req->buffer either...
-                        */
-                       req->buffer = bio_data(bio);
-                       req->current_nr_sectors = bio_cur_sectors(bio);
-                       req->hard_cur_sectors = req->current_nr_sectors;
-                       req->sector = req->hard_sector = bio->bi_sector;
-                       req->nr_sectors = req->hard_nr_sectors += nr_sectors;
-                       req->ioprio = ioprio_best(req->ioprio, prio);
-                       drive_stat_acct(req, 0);
-                       if (!attempt_front_merge(q, req))
-                               elv_merged_request(q, req, el_ret);
-                       goto out;
-
-               /* ELV_NO_MERGE: elevator says don't/can't merge. */
-               default:
-                       ;
+               /*
+                * may not be valid. if the low level driver said
+                * it didn't need a bounce buffer then it better
+                * not touch req->buffer either...
+                */
+               req->buffer = bio_data(bio);
+               req->current_nr_sectors = bio_cur_sectors(bio);
+               req->hard_cur_sectors = req->current_nr_sectors;
+               req->sector = req->hard_sector = bio->bi_sector;
+               req->nr_sectors = req->hard_nr_sectors += nr_sectors;
+               req->ioprio = ioprio_best(req->ioprio, prio);
+               drive_stat_acct(req, 0);
+               if (!attempt_front_merge(q, req))
+                       elv_merged_request(q, req, el_ret);
+               goto out;
+
+       /* ELV_NO_MERGE: elevator says don't/can't merge. */
+       default:
+               ;
        }
 
 get_rq:
@@ -1350,7 +1344,7 @@ end_io:
                }
 
                if (unlikely(nr_sectors > q->max_hw_sectors)) {
-                       printk("bio too big device %s (%u > %u)\n", 
+                       printk(KERN_ERR "bio too big device %s (%u > %u)\n",
                                bdevname(bio->bi_bdev, b),
                                bio_sectors(bio),
                                q->max_hw_sectors);
@@ -1439,7 +1433,6 @@ void generic_make_request(struct bio *bio)
        } while (bio);
        current->bio_tail = NULL; /* deactivate */
 }
-
 EXPORT_SYMBOL(generic_make_request);
 
 /**
@@ -1480,13 +1473,12 @@ void submit_bio(int rw, struct bio *bio)
                        current->comm, task_pid_nr(current),
                                (rw & WRITE) ? "WRITE" : "READ",
                                (unsigned long long)bio->bi_sector,
-                               bdevname(bio->bi_bdev,b));
+                               bdevname(bio->bi_bdev, b));
                }
        }
 
        generic_make_request(bio);
 }
-
 EXPORT_SYMBOL(submit_bio);
 
 /**
@@ -1518,9 +1510,8 @@ static int __end_that_request_first(struct request *req, int error,
        if (!blk_pc_request(req))
                req->errors = 0;
 
-       if (error) {
-               if (blk_fs_request(req) && !(req->cmd_flags & REQ_QUIET))
-                       printk("end_request: I/O error, dev %s, sector %llu\n",
+       if (error && (blk_fs_request(req) && !(req->cmd_flags & REQ_QUIET))) {
+               printk(KERN_ERR "end_request: I/O error, dev %s, sector %llu\n",
                                req->rq_disk ? req->rq_disk->disk_name : "?",
                                (unsigned long long)req->sector);
        }
@@ -1554,9 +1545,9 @@ static int __end_that_request_first(struct request *req, int error,
 
                        if (unlikely(bio->bi_idx >= bio->bi_vcnt)) {
                                blk_dump_rq_flags(req, "__end_that");
-                               printk("%s: bio idx %d >= vcnt %d\n",
-                                               __FUNCTION__,
-                                               bio->bi_idx, bio->bi_vcnt);
+                               printk(KERN_ERR "%s: bio idx %d >= vcnt %d\n",
+                                               __FUNCTION__, bio->bi_idx,
+                                               bio->bi_vcnt);
                                break;
                        }
 
@@ -1582,7 +1573,8 @@ static int __end_that_request_first(struct request *req, int error,
                total_bytes += nbytes;
                nr_bytes -= nbytes;
 
-               if ((bio = req->bio)) {
+               bio = req->bio;
+               if (bio) {
                        /*
                         * end more in this run, or just return 'not-done'
                         */
@@ -1626,15 +1618,16 @@ static void blk_done_softirq(struct softirq_action *h)
        local_irq_enable();
 
        while (!list_empty(&local_list)) {
-               struct request *rq = list_entry(local_list.next, struct request, donelist);
+               struct request *rq;
 
+               rq = list_entry(local_list.next, struct request, donelist);
                list_del_init(&rq->donelist);
                rq->q->softirq_done_fn(rq);
        }
 }
 
-static int __cpuinit blk_cpu_notify(struct notifier_block *self, unsigned long action,
-                         void *hcpu)
+static int __cpuinit blk_cpu_notify(struct notifier_block *self,
+                                   unsigned long action, void *hcpu)
 {
        /*
         * If a CPU goes away, splice its entries to the current CPU
@@ -1676,7 +1669,7 @@ void blk_complete_request(struct request *req)
        unsigned long flags;
 
        BUG_ON(!req->q->softirq_done_fn);
-               
+
        local_irq_save(flags);
 
        cpu_list = &__get_cpu_var(blk_cpu_done);
@@ -1685,9 +1678,8 @@ void blk_complete_request(struct request *req)
 
        local_irq_restore(flags);
 }
-
 EXPORT_SYMBOL(blk_complete_request);
-       
+
 /*
  * queue lock must be held
  */
@@ -2002,7 +1994,6 @@ int kblockd_schedule_work(struct work_struct *work)
 {
        return queue_work(kblockd_workqueue, work);
 }
-
 EXPORT_SYMBOL(kblockd_schedule_work);
 
 void kblockd_flush_work(struct work_struct *work)