On 8/12/23 16:06, Yu Kuai wrote:

Hi,

在 2023/08/08 11:32, Xueshi Hu 写道:

quoted

Because raid reshape changes the r1conf::raid_disks and the mempool, it
orders that there's no in-flight r1bio when reshaping. However, the
current caller of allow_barrier() allows the reshape
operation to proceed even if the old r1bio requests have not been freed.

Free the r1bio firstly before allow_barrier()

Fixes: c91114c2b89d ("md/raid1: release pending accounting for an I/O 
only after write-behind is also finished")
Fixes: 6bfe0b499082 ("md: support blocking writes to an array on 
device failure")
Fixes: 689389a06ce7 ("md/raid1: simplify handle_read_error().")

It's just a suggestion, you can split this patch into 3 patches, one for
each fix.

quoted

Reviewed-by: Yu Kuai <redacted>
Signed-off-by: Xueshi Hu <redacted>
---
-> v2:
    - fix the problem one by one instead of calling
    blk_mq_freeze_queue() as suggested by Yu Kuai
-> v3:
    - add freeze_array_totally() to replace freeze_array() instead
      of gave up in raid1_reshape()
    - add a missed fix in raid_end_bio_io()
    - add a small check at the start of raid1_reshape()
-> v4:
    - add fix tag and revise the commit message
    - drop patch 1 as there is an ongoing systematic fix for the bug

I think it's okay to fix the bug as I suggested in v3, the refactor is a
long term...

Thanks,
Kuai

Good suggestion, I'll remake the patch.

Thanks,
Hu

quoted

    - drop patch 3 as it's unrelated which will be sent in
    another patch

  drivers/md/raid1.c | 15 +++++++++------
  1 file changed, 9 insertions(+), 6 deletions(-)

diff --git a/drivers/md/raid1.c b/drivers/md/raid1.c
index dd25832eb045..5a5eb5f1a224 100644
--- a/drivers/md/raid1.c
+++ b/drivers/md/raid1.c

@@ -313,6 +313,7 @@ static void raid_end_bio_io(struct r1bio *r1_bio)

  {
      struct bio *bio = r1_bio->master_bio;
      struct r1conf *conf = r1_bio->mddev->private;
+    sector_t sector = r1_bio->sector;
      /* if nobody has done the final endio yet, do it now */
      if (!test_and_set_bit(R1BIO_Returned, &r1_bio->state)) {

@@ -323,13 +324,13 @@ static void raid_end_bio_io(struct r1bio *r1_bio)

          call_bio_endio(r1_bio);
      }
+
+    free_r1bio(r1_bio);
      /*
       * Wake up any possible resync thread that waits for the device
       * to go idle.  All I/Os, even write-behind writes, are done.
       */
-    allow_barrier(conf, r1_bio->sector);
-
-    free_r1bio(r1_bio);
+    allow_barrier(conf, sector);
  }
  /*

@@ -1373,6 +1374,7 @@ static void raid1_write_request(struct mddev 

*mddev, struct bio *bio,
          return;
      }
+ retry_write:
      r1_bio = alloc_r1bio(mddev, bio);
      r1_bio->sectors = max_write_sectors;

@@ -1388,7 +1390,6 @@ static void raid1_write_request(struct mddev 

*mddev, struct bio *bio,
       */
      disks = conf->raid_disks * 2;
- retry_write:
      blocked_rdev = NULL;
      rcu_read_lock();
      max_sectors = r1_bio->sectors;

@@ -1468,7 +1469,7 @@ static void raid1_write_request(struct mddev 

*mddev, struct bio *bio,
          for (j = 0; j < i; j++)
              if (r1_bio->bios[j])
                  rdev_dec_pending(conf->mirrors[j].rdev, mddev);
-        r1_bio->state = 0;
+        free_r1bio(r1_bio);
          allow_barrier(conf, bio->bi_iter.bi_sector);
          if (bio->bi_opf & REQ_NOWAIT) {

@@ -2498,6 +2499,7 @@ static void handle_read_error(struct r1conf 

*conf, struct r1bio *r1_bio)
      struct mddev *mddev = conf->mddev;
      struct bio *bio;
      struct md_rdev *rdev;
+    sector_t sector;
      clear_bit(R1BIO_ReadError, &r1_bio->state);
      /* we got a read error. Maybe the drive is bad.  Maybe just

@@ -2527,12 +2529,13 @@ static void handle_read_error(struct r1conf 

*conf, struct r1bio *r1_bio)
      }
      rdev_dec_pending(rdev, conf->mddev);
-    allow_barrier(conf, r1_bio->sector);
+    sector = r1_bio->sector;
      bio = r1_bio->master_bio;
      /* Reuse the old r1_bio so that the IO_BLOCKED settings are 
preserved */
      r1_bio->state = 0;
      raid1_read_request(mddev, bio, r1_bio->sectors, r1_bio);
+    allow_barrier(conf, sector);
  }
  static void raid1d(struct md_thread *thread)