[BLOCK] ll_rw_blk: Enable out-of-order request completions through softirq
Request completion can be a quite heavy process, since it needs to
iterate through the entire request and complete the bio's it holds.
This patch adds blk_complete_request() which moves this processing
into a dedicated block softirq.
Signed-off-by: Jens Axboe <axboe@suse.de>
diff --git a/block/ll_rw_blk.c b/block/ll_rw_blk.c
index 91d3b48..8e13645 100644
--- a/block/ll_rw_blk.c
+++ b/block/ll_rw_blk.c
@@ -27,6 +27,8 @@
#include <linux/swap.h>
#include <linux/writeback.h>
#include <linux/blkdev.h>
+#include <linux/interrupt.h>
+#include <linux/cpu.h>
/*
* for max sense size
@@ -62,13 +64,15 @@
/*
* Controlling structure to kblockd
*/
-static struct workqueue_struct *kblockd_workqueue;
+static struct workqueue_struct *kblockd_workqueue;
unsigned long blk_max_low_pfn, blk_max_pfn;
EXPORT_SYMBOL(blk_max_low_pfn);
EXPORT_SYMBOL(blk_max_pfn);
+static DEFINE_PER_CPU(struct list_head, blk_cpu_done);
+
/* Amount of time in which a process may batch requests */
#define BLK_BATCH_TIME (HZ/50UL)
@@ -207,6 +211,13 @@
EXPORT_SYMBOL(blk_queue_merge_bvec);
+void blk_queue_softirq_done(request_queue_t *q, softirq_done_fn *fn)
+{
+ q->softirq_done_fn = fn;
+}
+
+EXPORT_SYMBOL(blk_queue_softirq_done);
+
/**
* blk_queue_make_request - define an alternate make_request function for a device
* @q: the request queue for the device to be affected
@@ -270,6 +281,7 @@
static inline void rq_init(request_queue_t *q, struct request *rq)
{
INIT_LIST_HEAD(&rq->queuelist);
+ INIT_LIST_HEAD(&rq->donelist);
rq->errors = 0;
rq->rq_status = RQ_ACTIVE;
@@ -286,6 +298,7 @@
rq->sense = NULL;
rq->end_io = NULL;
rq->end_io_data = NULL;
+ rq->completion_data = NULL;
}
/**
@@ -3287,6 +3300,87 @@
EXPORT_SYMBOL(end_that_request_chunk);
/*
+ * splice the completion data to a local structure and hand off to
+ * process_completion_queue() to complete the requests
+ */
+static void blk_done_softirq(struct softirq_action *h)
+{
+ struct list_head *cpu_list;
+ LIST_HEAD(local_list);
+
+ local_irq_disable();
+ cpu_list = &__get_cpu_var(blk_cpu_done);
+ list_splice_init(cpu_list, &local_list);
+ local_irq_enable();
+
+ while (!list_empty(&local_list)) {
+ struct request *rq = list_entry(local_list.next, struct request, donelist);
+
+ list_del_init(&rq->donelist);
+ rq->q->softirq_done_fn(rq);
+ }
+}
+
+#ifdef CONFIG_HOTPLUG_CPU
+
+static int blk_cpu_notify(struct notifier_block *self, unsigned long action,
+ void *hcpu)
+{
+ /*
+ * If a CPU goes away, splice its entries to the current CPU
+ * and trigger a run of the softirq
+ */
+ if (action == CPU_DEAD) {
+ int cpu = (unsigned long) hcpu;
+
+ local_irq_disable();
+ list_splice_init(&per_cpu(blk_cpu_done, cpu),
+ &__get_cpu_var(blk_cpu_done));
+ raise_softirq_irqoff(BLOCK_SOFTIRQ);
+ local_irq_enable();
+ }
+
+ return NOTIFY_OK;
+}
+
+
+static struct notifier_block __devinitdata blk_cpu_notifier = {
+ .notifier_call = blk_cpu_notify,
+};
+
+#endif /* CONFIG_HOTPLUG_CPU */
+
+/**
+ * blk_complete_request - end I/O on a request
+ * @req: the request being processed
+ *
+ * Description:
+ * Ends all I/O on a request. It does not handle partial completions,
+ * unless the driver actually implements this in its completionc callback
+ * through requeueing. Theh actual completion happens out-of-order,
+ * through a softirq handler. The user must have registered a completion
+ * callback through blk_queue_softirq_done().
+ **/
+
+void blk_complete_request(struct request *req)
+{
+ struct list_head *cpu_list;
+ unsigned long flags;
+
+ BUG_ON(!req->q->softirq_done_fn);
+
+ local_irq_save(flags);
+
+ cpu_list = &__get_cpu_var(blk_cpu_done);
+ list_add_tail(&req->donelist, cpu_list);
+ raise_softirq_irqoff(BLOCK_SOFTIRQ);
+
+ local_irq_restore(flags);
+}
+
+EXPORT_SYMBOL(blk_complete_request);
+
+/*
* queue lock must be held
*/
void end_that_request_last(struct request *req, int uptodate)
@@ -3364,6 +3458,8 @@
int __init blk_dev_init(void)
{
+ int i;
+
kblockd_workqueue = create_workqueue("kblockd");
if (!kblockd_workqueue)
panic("Failed to create kblockd\n");
@@ -3377,6 +3473,14 @@
iocontext_cachep = kmem_cache_create("blkdev_ioc",
sizeof(struct io_context), 0, SLAB_PANIC, NULL, NULL);
+ for (i = 0; i < NR_CPUS; i++)
+ INIT_LIST_HEAD(&per_cpu(blk_cpu_done, i));
+
+ open_softirq(BLOCK_SOFTIRQ, blk_done_softirq, NULL);
+#ifdef CONFIG_HOTPLUG_CPU
+ register_cpu_notifier(&blk_cpu_notifier);
+#endif
+
blk_max_low_pfn = max_low_pfn;
blk_max_pfn = max_pfn;
diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
index fb09853..804cc4e 100644
--- a/include/linux/blkdev.h
+++ b/include/linux/blkdev.h
@@ -118,9 +118,9 @@
* try to put the fields that are referenced together in the same cacheline
*/
struct request {
- struct list_head queuelist; /* looking for ->queue? you must _not_
- * access it directly, use
- * blkdev_dequeue_request! */
+ struct list_head queuelist;
+ struct list_head donelist;
+
unsigned long flags; /* see REQ_ bits below */
/* Maintain bio traversal state for part by part I/O submission.
@@ -141,6 +141,7 @@
struct bio *biotail;
void *elevator_private;
+ void *completion_data;
unsigned short ioprio;
@@ -291,6 +292,7 @@
typedef void (activity_fn) (void *data, int rw);
typedef int (issue_flush_fn) (request_queue_t *, struct gendisk *, sector_t *);
typedef void (prepare_flush_fn) (request_queue_t *, struct request *);
+typedef void (softirq_done_fn)(struct request *);
enum blk_queue_state {
Queue_down,
@@ -332,6 +334,7 @@
activity_fn *activity_fn;
issue_flush_fn *issue_flush_fn;
prepare_flush_fn *prepare_flush_fn;
+ softirq_done_fn *softirq_done_fn;
/*
* Dispatch queue sorting
@@ -646,6 +649,17 @@
extern int end_that_request_chunk(struct request *, int, int);
extern void end_that_request_last(struct request *, int);
extern void end_request(struct request *req, int uptodate);
+extern void blk_complete_request(struct request *);
+
+static inline int rq_all_done(struct request *rq, unsigned int nr_bytes)
+{
+ if (blk_fs_request(rq))
+ return (nr_bytes >= (rq->hard_nr_sectors << 9));
+ else if (blk_pc_request(rq))
+ return nr_bytes >= rq->data_len;
+
+ return 0;
+}
/*
* end_that_request_first/chunk() takes an uptodate argument. we account
@@ -694,6 +708,7 @@
extern void blk_queue_prep_rq(request_queue_t *, prep_rq_fn *pfn);
extern void blk_queue_merge_bvec(request_queue_t *, merge_bvec_fn *);
extern void blk_queue_dma_alignment(request_queue_t *, int);
+extern void blk_queue_softirq_done(request_queue_t *, softirq_done_fn *);
extern struct backing_dev_info *blk_get_backing_dev_info(struct block_device *bdev);
extern int blk_queue_ordered(request_queue_t *, unsigned, prepare_flush_fn *);
extern void blk_queue_issue_flush_fn(request_queue_t *, issue_flush_fn *);
diff --git a/include/linux/interrupt.h b/include/linux/interrupt.h
index e50a95f..f022047 100644
--- a/include/linux/interrupt.h
+++ b/include/linux/interrupt.h
@@ -112,6 +112,7 @@
TIMER_SOFTIRQ,
NET_TX_SOFTIRQ,
NET_RX_SOFTIRQ,
+ BLOCK_SOFTIRQ,
SCSI_SOFTIRQ,
TASKLET_SOFTIRQ
};