Thread (23 messages) 23 messages, 4 authors, 2024-03-01

Re: [PATCH RFC 3/4] md: Missing decrease active_io for flush io

From: Xiao Ni <hidden>
Date: 2024-02-23 13:49:56
Also in: dm-devel

On Fri, Feb 23, 2024 at 11:06 AM Yu Kuai [off-list ref] wrote:
Hi,

在 2024/02/20 23:30, Xiao Ni 写道:
quoted
If all flush bios finish fast, it doesn't decrease active_io. And it will
stuck when stopping array.

This can be reproduced by lvm2 test shell/integrity-caching.sh.
But it can't reproduce 100%.

Fixes: fa2bbff7b0b4 ("md: synchronize flush io with array reconfiguration")
Signed-off-by: Xiao Ni <redacted>
Same patch is already applied:

855678ed8534 md: Fix missing release of 'active_io' for flush
Thanks for the information. I used Linus's tree so I missed this one.

Regards
Xiao
Thanks,
Kuai
quoted
---
  drivers/md/md.c | 6 +++++-
  1 file changed, 5 insertions(+), 1 deletion(-)
diff --git a/drivers/md/md.c b/drivers/md/md.c
index 77e3af7cf6bb..a41bed286fd2 100644
--- a/drivers/md/md.c
+++ b/drivers/md/md.c
@@ -579,8 +579,12 @@ static void submit_flushes(struct work_struct *ws)
                      rcu_read_lock();
              }
      rcu_read_unlock();
-     if (atomic_dec_and_test(&mddev->flush_pending))
+     if (atomic_dec_and_test(&mddev->flush_pending)) {
+             /* The pair is percpu_ref_get() from md_flush_request() */
+             percpu_ref_put(&mddev->active_io);
+
              queue_work(md_wq, &mddev->flush_work);
+     }
  }

  static void md_submit_flush_data(struct work_struct *ws)
  
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help