Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / block / blk-timeout.c
diff --git a/kernel/block/blk-timeout.c b/kernel/block/blk-timeout.c
new file mode 100644 (file)
index 0000000..246dfb1
--- /dev/null
@@ -0,0 +1,235 @@
+/*
+ * Functions related to generic timeout handling of requests.
+ */
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/blkdev.h>
+#include <linux/fault-inject.h>
+
+#include "blk.h"
+#include "blk-mq.h"
+
+#ifdef CONFIG_FAIL_IO_TIMEOUT
+
+static DECLARE_FAULT_ATTR(fail_io_timeout);
+
+static int __init setup_fail_io_timeout(char *str)
+{
+       return setup_fault_attr(&fail_io_timeout, str);
+}
+__setup("fail_io_timeout=", setup_fail_io_timeout);
+
+int blk_should_fake_timeout(struct request_queue *q)
+{
+       if (!test_bit(QUEUE_FLAG_FAIL_IO, &q->queue_flags))
+               return 0;
+
+       return should_fail(&fail_io_timeout, 1);
+}
+
+static int __init fail_io_timeout_debugfs(void)
+{
+       struct dentry *dir = fault_create_debugfs_attr("fail_io_timeout",
+                                               NULL, &fail_io_timeout);
+
+       return PTR_ERR_OR_ZERO(dir);
+}
+
+late_initcall(fail_io_timeout_debugfs);
+
+ssize_t part_timeout_show(struct device *dev, struct device_attribute *attr,
+                         char *buf)
+{
+       struct gendisk *disk = dev_to_disk(dev);
+       int set = test_bit(QUEUE_FLAG_FAIL_IO, &disk->queue->queue_flags);
+
+       return sprintf(buf, "%d\n", set != 0);
+}
+
+ssize_t part_timeout_store(struct device *dev, struct device_attribute *attr,
+                          const char *buf, size_t count)
+{
+       struct gendisk *disk = dev_to_disk(dev);
+       int val;
+
+       if (count) {
+               struct request_queue *q = disk->queue;
+               char *p = (char *) buf;
+
+               val = simple_strtoul(p, &p, 10);
+               spin_lock_irq(q->queue_lock);
+               if (val)
+                       queue_flag_set(QUEUE_FLAG_FAIL_IO, q);
+               else
+                       queue_flag_clear(QUEUE_FLAG_FAIL_IO, q);
+               spin_unlock_irq(q->queue_lock);
+       }
+
+       return count;
+}
+
+#endif /* CONFIG_FAIL_IO_TIMEOUT */
+
+/*
+ * blk_delete_timer - Delete/cancel timer for a given function.
+ * @req:       request that we are canceling timer for
+ *
+ */
+void blk_delete_timer(struct request *req)
+{
+       list_del_init(&req->timeout_list);
+}
+
+static void blk_rq_timed_out(struct request *req)
+{
+       struct request_queue *q = req->q;
+       enum blk_eh_timer_return ret = BLK_EH_RESET_TIMER;
+
+       if (q->rq_timed_out_fn)
+               ret = q->rq_timed_out_fn(req);
+       switch (ret) {
+       case BLK_EH_HANDLED:
+               /* Can we use req->errors here? */
+               __blk_complete_request(req);
+               break;
+       case BLK_EH_RESET_TIMER:
+               blk_add_timer(req);
+               blk_clear_rq_complete(req);
+               break;
+       case BLK_EH_NOT_HANDLED:
+               /*
+                * LLD handles this for now but in the future
+                * we can send a request msg to abort the command
+                * and we can move more of the generic scsi eh code to
+                * the blk layer.
+                */
+               break;
+       default:
+               printk(KERN_ERR "block: bad eh return: %d\n", ret);
+               break;
+       }
+}
+
+static void blk_rq_check_expired(struct request *rq, unsigned long *next_timeout,
+                         unsigned int *next_set)
+{
+       if (time_after_eq(jiffies, rq->deadline)) {
+               list_del_init(&rq->timeout_list);
+
+               /*
+                * Check if we raced with end io completion
+                */
+               if (!blk_mark_rq_complete(rq))
+                       blk_rq_timed_out(rq);
+       } else if (!*next_set || time_after(*next_timeout, rq->deadline)) {
+               *next_timeout = rq->deadline;
+               *next_set = 1;
+       }
+}
+
+void blk_rq_timed_out_timer(unsigned long data)
+{
+       struct request_queue *q = (struct request_queue *) data;
+       unsigned long flags, next = 0;
+       struct request *rq, *tmp;
+       int next_set = 0;
+
+       spin_lock_irqsave(q->queue_lock, flags);
+
+       list_for_each_entry_safe(rq, tmp, &q->timeout_list, timeout_list)
+               blk_rq_check_expired(rq, &next, &next_set);
+
+       if (next_set)
+               mod_timer(&q->timeout, round_jiffies_up(next));
+
+       spin_unlock_irqrestore(q->queue_lock, flags);
+}
+
+/**
+ * blk_abort_request -- Request request recovery for the specified command
+ * @req:       pointer to the request of interest
+ *
+ * This function requests that the block layer start recovery for the
+ * request by deleting the timer and calling the q's timeout function.
+ * LLDDs who implement their own error recovery MAY ignore the timeout
+ * event if they generated blk_abort_req. Must hold queue lock.
+ */
+void blk_abort_request(struct request *req)
+{
+       if (blk_mark_rq_complete(req))
+               return;
+       blk_delete_timer(req);
+       if (req->q->mq_ops)
+               blk_mq_rq_timed_out(req, false);
+       else
+               blk_rq_timed_out(req);
+}
+EXPORT_SYMBOL_GPL(blk_abort_request);
+
+unsigned long blk_rq_timeout(unsigned long timeout)
+{
+       unsigned long maxt;
+
+       maxt = round_jiffies_up(jiffies + BLK_MAX_TIMEOUT);
+       if (time_after(timeout, maxt))
+               timeout = maxt;
+
+       return timeout;
+}
+
+/**
+ * blk_add_timer - Start timeout timer for a single request
+ * @req:       request that is about to start running.
+ *
+ * Notes:
+ *    Each request has its own timer, and as it is added to the queue, we
+ *    set up the timer. When the request completes, we cancel the timer.
+ */
+void blk_add_timer(struct request *req)
+{
+       struct request_queue *q = req->q;
+       unsigned long expiry;
+
+       if (req->cmd_flags & REQ_NO_TIMEOUT)
+               return;
+
+       /* blk-mq has its own handler, so we don't need ->rq_timed_out_fn */
+       if (!q->mq_ops && !q->rq_timed_out_fn)
+               return;
+
+       BUG_ON(!list_empty(&req->timeout_list));
+
+       /*
+        * Some LLDs, like scsi, peek at the timeout to prevent a
+        * command from being retried forever.
+        */
+       if (!req->timeout)
+               req->timeout = q->rq_timeout;
+
+       req->deadline = jiffies + req->timeout;
+       if (!q->mq_ops)
+               list_add_tail(&req->timeout_list, &req->q->timeout_list);
+
+       /*
+        * If the timer isn't already pending or this timeout is earlier
+        * than an existing one, modify the timer. Round up to next nearest
+        * second.
+        */
+       expiry = blk_rq_timeout(round_jiffies_up(req->deadline));
+
+       if (!timer_pending(&q->timeout) ||
+           time_before(expiry, q->timeout.expires)) {
+               unsigned long diff = q->timeout.expires - expiry;
+
+               /*
+                * Due to added timer slack to group timers, the timer
+                * will often be a little in front of what we asked for.
+                * So apply some tolerance here too, otherwise we keep
+                * modifying the timer because expires for value X
+                * will be X + something.
+                */
+               if (!timer_pending(&q->timeout) || (diff >= HZ / 2))
+                       mod_timer(&q->timeout, expiry);
+       }
+
+}