md/raid5: don't complete make_request on barrier until writes are scheduled

The post-barrier-flush is sent by md as soon as make_request on the
barrier write completes.  For raid5, the data might not be in the
per-device queues yet.  So for barrier requests, wait for any
pre-reading to be done so that the request will be in the per-device
queues.

We use the 'preread_active' count to check that nothing is still in
the preread phase, and delay the decrement of this count until after
write requests have been submitted to the underlying devices.

Signed-off-by: NeilBrown <neilb@suse.de>

diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c
index 8c9395f2028f54faee1e65d6b663bcb053e67991..c78d4835b0d6df400da982f5cf58b2cb1050e3a2 100644 (file)
--- a/drivers/md/raid5.c
+++ b/drivers/md/raid5.c
@@ -2947,6 +2947,7 @@ static void handle_stripe5(struct stripe_head *sh)
        struct r5dev *dev;
        mdk_rdev_t *blocked_rdev = NULL;
        int prexor;
+       int dec_preread_active = 0;
 
        memset(&s, 0, sizeof(s));
        pr_debug("handling stripe %llu, state=%#lx cnt=%d, pd_idx=%d check:%d "
@@ -3096,12 +3097,8 @@ static void handle_stripe5(struct stripe_head *sh)
                                        set_bit(STRIPE_INSYNC, &sh->state);
                        }
                }
-               if (test_and_clear_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) {
-                       atomic_dec(&conf->preread_active_stripes);
-                       if (atomic_read(&conf->preread_active_stripes) <
-                               IO_THRESHOLD)
-                               md_wakeup_thread(conf->mddev->thread);
-               }
+               if (test_and_clear_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
+                       dec_preread_active = 1;
        }
 
        /* Now to consider new write requests and what else, if anything
@@ -3208,6 +3205,16 @@ static void handle_stripe5(struct stripe_head *sh)
 
        ops_run_io(sh, &s);
 
+       if (dec_preread_active) {
+               /* We delay this until after ops_run_io so that if make_request
+                * is waiting on a barrier, it won't continue until the writes
+                * have actually been submitted.
+                */
+               atomic_dec(&conf->preread_active_stripes);
+               if (atomic_read(&conf->preread_active_stripes) <
+                   IO_THRESHOLD)
+                       md_wakeup_thread(conf->mddev->thread);
+       }
        return_io(return_bi);
 }
 
@@ -3221,6 +3228,7 @@ static void handle_stripe6(struct stripe_head *sh)
        struct r6_state r6s;
        struct r5dev *dev, *pdev, *qdev;
        mdk_rdev_t *blocked_rdev = NULL;
+       int dec_preread_active = 0;
 
        pr_debug("handling stripe %llu, state=%#lx cnt=%d, "
                "pd_idx=%d, qd_idx=%d\n, check:%d, reconstruct:%d\n",
@@ -3379,12 +3387,8 @@ static void handle_stripe6(struct stripe_head *sh)
                                        set_bit(STRIPE_INSYNC, &sh->state);
                        }
                }
-               if (test_and_clear_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) {
-                       atomic_dec(&conf->preread_active_stripes);
-                       if (atomic_read(&conf->preread_active_stripes) <
-                               IO_THRESHOLD)
-                               md_wakeup_thread(conf->mddev->thread);
-               }
+               if (test_and_clear_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
+                       dec_preread_active = 1;
        }
 
        /* Now to consider new write requests and what else, if anything
@@ -3493,6 +3497,18 @@ static void handle_stripe6(struct stripe_head *sh)
 
        ops_run_io(sh, &s);
 
+
+       if (dec_preread_active) {
+               /* We delay this until after ops_run_io so that if make_request
+                * is waiting on a barrier, it won't continue until the writes
+                * have actually been submitted.
+                */
+               atomic_dec(&conf->preread_active_stripes);
+               if (atomic_read(&conf->preread_active_stripes) <
+                   IO_THRESHOLD)
+                       md_wakeup_thread(conf->mddev->thread);
+       }
+
        return_io(return_bi);
 }
 
@@ -3995,6 +4011,9 @@ static int make_request(struct request_queue *q, struct bio * bi)
                        finish_wait(&conf->wait_for_overlap, &w);
                        set_bit(STRIPE_HANDLE, &sh->state);
                        clear_bit(STRIPE_DELAYED, &sh->state);
+                       if (mddev->barrier && 
+                           !test_and_set_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
+                               atomic_inc(&conf->preread_active_stripes);
                        release_stripe(sh);
                } else {
                        /* cannot get stripe for read-ahead, just give-up */
@@ -4014,6 +4033,14 @@ static int make_request(struct request_queue *q, struct bio * bi)
 
                bio_endio(bi, 0);
        }
+
+       if (mddev->barrier) {
+               /* We need to wait for the stripes to all be handled.
+                * So: wait for preread_active_stripes to drop to 0.
+                */
+               wait_event(mddev->thread->wqueue,
+                          atomic_read(&conf->preread_active_stripes) == 0);
+       }
        return 0;
 }