summaryrefslogtreecommitdiff
path: root/block/blk-ioc.c
diff options
context:
space:
mode:
authorJens Axboe <jens.axboe@oracle.com>2008-01-29 14:53:40 +0100
committerJens Axboe <jens.axboe@oracle.com>2008-01-29 21:55:08 +0100
commit86db1e29772372155db08ff48a9ceb76e11a2ad1 (patch)
tree312f38eb3245873c476c50f816b85610fef9615a /block/blk-ioc.c
parent8324aa91d1e11a1fc25f209687a0b2e6c2ed47d0 (diff)
downloadlwn-86db1e29772372155db08ff48a9ceb76e11a2ad1.tar.gz
lwn-86db1e29772372155db08ff48a9ceb76e11a2ad1.zip
block: continue ll_rw_blk.c splitup
Adds files for barrier handling, rq execution, io context handling, mapping data to requests, and queue settings. Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Diffstat (limited to 'block/blk-ioc.c')
-rw-r--r--block/blk-ioc.c194
1 files changed, 194 insertions, 0 deletions
diff --git a/block/blk-ioc.c b/block/blk-ioc.c
new file mode 100644
index 000000000000..6d1675508eb5
--- /dev/null
+++ b/block/blk-ioc.c
@@ -0,0 +1,194 @@
+/*
+ * Functions related to io context handling
+ */
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/bio.h>
+#include <linux/blkdev.h>
+#include <linux/bootmem.h> /* for max_pfn/max_low_pfn */
+
+#include "blk.h"
+
+/*
+ * For io context allocations
+ */
+static struct kmem_cache *iocontext_cachep;
+
+static void cfq_dtor(struct io_context *ioc)
+{
+ struct cfq_io_context *cic[1];
+ int r;
+
+ /*
+ * We don't have a specific key to lookup with, so use the gang
+ * lookup to just retrieve the first item stored. The cfq exit
+ * function will iterate the full tree, so any member will do.
+ */
+ r = radix_tree_gang_lookup(&ioc->radix_root, (void **) cic, 0, 1);
+ if (r > 0)
+ cic[0]->dtor(ioc);
+}
+
+/*
+ * IO Context helper functions. put_io_context() returns 1 if there are no
+ * more users of this io context, 0 otherwise.
+ */
+int put_io_context(struct io_context *ioc)
+{
+ if (ioc == NULL)
+ return 1;
+
+ BUG_ON(atomic_read(&ioc->refcount) == 0);
+
+ if (atomic_dec_and_test(&ioc->refcount)) {
+ rcu_read_lock();
+ if (ioc->aic && ioc->aic->dtor)
+ ioc->aic->dtor(ioc->aic);
+ rcu_read_unlock();
+ cfq_dtor(ioc);
+
+ kmem_cache_free(iocontext_cachep, ioc);
+ return 1;
+ }
+ return 0;
+}
+EXPORT_SYMBOL(put_io_context);
+
+static void cfq_exit(struct io_context *ioc)
+{
+ struct cfq_io_context *cic[1];
+ int r;
+
+ rcu_read_lock();
+ /*
+ * See comment for cfq_dtor()
+ */
+ r = radix_tree_gang_lookup(&ioc->radix_root, (void **) cic, 0, 1);
+ rcu_read_unlock();
+
+ if (r > 0)
+ cic[0]->exit(ioc);
+}
+
+/* Called by the exitting task */
+void exit_io_context(void)
+{
+ struct io_context *ioc;
+
+ task_lock(current);
+ ioc = current->io_context;
+ current->io_context = NULL;
+ task_unlock(current);
+
+ if (atomic_dec_and_test(&ioc->nr_tasks)) {
+ if (ioc->aic && ioc->aic->exit)
+ ioc->aic->exit(ioc->aic);
+ cfq_exit(ioc);
+
+ put_io_context(ioc);
+ }
+}
+
+struct io_context *alloc_io_context(gfp_t gfp_flags, int node)
+{
+ struct io_context *ret;
+
+ ret = kmem_cache_alloc_node(iocontext_cachep, gfp_flags, node);
+ if (ret) {
+ atomic_set(&ret->refcount, 1);
+ atomic_set(&ret->nr_tasks, 1);
+ spin_lock_init(&ret->lock);
+ ret->ioprio_changed = 0;
+ ret->ioprio = 0;
+ ret->last_waited = jiffies; /* doesn't matter... */
+ ret->nr_batch_requests = 0; /* because this is 0 */
+ ret->aic = NULL;
+ INIT_RADIX_TREE(&ret->radix_root, GFP_ATOMIC | __GFP_HIGH);
+ ret->ioc_data = NULL;
+ }
+
+ return ret;
+}
+
+/*
+ * If the current task has no IO context then create one and initialise it.
+ * Otherwise, return its existing IO context.
+ *
+ * This returned IO context doesn't have a specifically elevated refcount,
+ * but since the current task itself holds a reference, the context can be
+ * used in general code, so long as it stays within `current` context.
+ */
+struct io_context *current_io_context(gfp_t gfp_flags, int node)
+{
+ struct task_struct *tsk = current;
+ struct io_context *ret;
+
+ ret = tsk->io_context;
+ if (likely(ret))
+ return ret;
+
+ ret = alloc_io_context(gfp_flags, node);
+ if (ret) {
+ /* make sure set_task_ioprio() sees the settings above */
+ smp_wmb();
+ tsk->io_context = ret;
+ }
+
+ return ret;
+}
+
+/*
+ * If the current task has no IO context then create one and initialise it.
+ * If it does have a context, take a ref on it.
+ *
+ * This is always called in the context of the task which submitted the I/O.
+ */
+struct io_context *get_io_context(gfp_t gfp_flags, int node)
+{
+ struct io_context *ret = NULL;
+
+ /*
+ * Check for unlikely race with exiting task. ioc ref count is
+ * zero when ioc is being detached.
+ */
+ do {
+ ret = current_io_context(gfp_flags, node);
+ if (unlikely(!ret))
+ break;
+ } while (!atomic_inc_not_zero(&ret->refcount));
+
+ return ret;
+}
+EXPORT_SYMBOL(get_io_context);
+
+void copy_io_context(struct io_context **pdst, struct io_context **psrc)
+{
+ struct io_context *src = *psrc;
+ struct io_context *dst = *pdst;
+
+ if (src) {
+ BUG_ON(atomic_read(&src->refcount) == 0);
+ atomic_inc(&src->refcount);
+ put_io_context(dst);
+ *pdst = src;
+ }
+}
+EXPORT_SYMBOL(copy_io_context);
+
+void swap_io_context(struct io_context **ioc1, struct io_context **ioc2)
+{
+ struct io_context *temp;
+ temp = *ioc1;
+ *ioc1 = *ioc2;
+ *ioc2 = temp;
+}
+EXPORT_SYMBOL(swap_io_context);
+
+int __init blk_ioc_init(void)
+{
+ iocontext_cachep = kmem_cache_create("blkdev_ioc",
+ sizeof(struct io_context), 0, SLAB_PANIC, NULL);
+ return 0;
+}
+subsys_initcall(blk_ioc_init);