lkml.org 
[lkml]   [2014]   [Jun]   [30]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
From
Subject[PATCH 04/18] drbd: reduce number of spinlock drop/re-aquire cycles
Date
From: Lars Ellenberg <lars.ellenberg@linbit.com>

Instead of dropping and re-aquiring the spinlock around the submit,
just remember that we want to submit, and do that only once we have
dropped the spinlock for good.

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
---
drivers/block/drbd/drbd_req.c | 20 ++++++++++++++------
1 file changed, 14 insertions(+), 6 deletions(-)

diff --git a/drivers/block/drbd/drbd_req.c b/drivers/block/drbd/drbd_req.c
index 09803d0..4c7fee1 100644
--- a/drivers/block/drbd/drbd_req.c
+++ b/drivers/block/drbd/drbd_req.c
@@ -1086,11 +1086,13 @@ drbd_request_prepare(struct drbd_device *device, struct bio *bio, unsigned long

static void drbd_send_and_submit(struct drbd_device *device, struct drbd_request *req)
{
+ struct drbd_resource *resource = device->resource;
const int rw = bio_rw(req->master_bio);
struct bio_and_error m = { NULL, };
bool no_remote = false;
+ bool submit_private_bio = false;

- spin_lock_irq(&device->resource->req_lock);
+ spin_lock_irq(&resource->req_lock);
if (rw == WRITE) {
/* This may temporarily give up the req_lock,
* but will re-aquire it before it returns here.
@@ -1152,9 +1154,7 @@ static void drbd_send_and_submit(struct drbd_device *device, struct drbd_request
/* needs to be marked within the same spinlock */
_req_mod(req, TO_BE_SUBMITTED);
/* but we need to give up the spinlock to submit */
- spin_unlock_irq(&device->resource->req_lock);
- drbd_submit_req_private_bio(req);
- spin_lock_irq(&device->resource->req_lock);
+ submit_private_bio = true;
} else if (no_remote) {
nodata:
if (__ratelimit(&drbd_ratelimit_state))
@@ -1167,8 +1167,16 @@ nodata:
out:
if (drbd_req_put_completion_ref(req, &m, 1))
kref_put(&req->kref, drbd_req_destroy);
- spin_unlock_irq(&device->resource->req_lock);
-
+ spin_unlock_irq(&resource->req_lock);
+
+ /* Even though above is a kref_put(), this is safe.
+ * As long as we still need to submit our private bio,
+ * we hold a completion ref, and the request cannot disappear.
+ * If however this request did not even have a private bio to submit
+ * (e.g. remote read), req may already be invalid now.
+ * That's why we cannot check on req->private_bio. */
+ if (submit_private_bio)
+ drbd_submit_req_private_bio(req);
if (m.bio)
complete_master_bio(device, &m);
}
--
1.9.1


\
 
 \ /
  Last update: 2014-06-30 18:41    [W:0.128 / U:5.576 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site