All of lore.kernel.org
 help / color / mirror / Atom feed
From: paulmck@linux.vnet.ibm.com (Paul E. McKenney)
Subject: [PATCH] nvme: avoid flush dependency in delete controller flow
Date: Mon, 9 Apr 2018 09:48:08 -0700	[thread overview]
Message-ID: <20180409164808.GK3948@linux.vnet.ibm.com> (raw)
In-Reply-To: <1523285426-26526-1-git-send-email-nitzanc@mellanox.com>

On Mon, Apr 09, 2018@05:50:26PM +0300, Nitzan Carmi wrote:
> nvme_delete_ctrl queues a work on a MEM_RECLAIM queue
> (nvme_delete_wq), which eventually calls cleanup_srcu_struct,
> which in turn flushes a delayed work from an !MEM_RECLAIM
> queue. This is unsafe as we might trigger deadlocks under
> severe memory pressure.
> 
> Since we don't ever invoke call_srcu(), it is safe
> to use the _quiesced() version of srcu cleanup, and
> avoid that flush dependency.
> 
> Signed-off-by: Nitzan Carmi <nitzanc at mellanox.com>

Very good, thank you!

I queued this with a few edits to the commit log (please see below),
so please let me know if I messed anything up.

Given that it only happens under high memory pressure, I am assuming
that sending this up the v4.18 merge window (not this one, but the next
one) will work.  Please let me know if this is more urgent than that.
(If it is not more urgent, I would like to allow the extra time for
people to get used to yet another addition to the SRCU API.)

------------------------------------------------------------------------

commit 6fde861519fc6e4be50c130272be556ef5656324
Author: Nitzan Carmi <nitzanc at mellanox.com>
Date:   Mon Apr 9 17:50:26 2018 +0300

    nvme: Avoid flush dependency in delete controller flow
    
    The nvme_delete_ctrl() function queues a work item on a MEM_RECLAIM
    queue (nvme_delete_wq), which eventually calls cleanup_srcu_struct(),
    which in turn flushes a delayed work from an !MEM_RECLAIM queue. This
    is unsafe as we might trigger deadlocks under severe memory pressure.
    
    Since we don't ever invoke call_srcu(), it is safe to use the shiny new
    _quiesced() version of srcu cleanup, thus avoiding that flush dependency.
    This commit makes that change.
    
    Signed-off-by: Nitzan Carmi <nitzanc at mellanox.com>
    Signed-off-by: Paul E. McKenney <paulmck at linux.vnet.ibm.com>

diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c
index 7aeca5db7916..40008d06c365 100644
--- a/drivers/nvme/host/core.c
+++ b/drivers/nvme/host/core.c
@@ -351,7 +351,7 @@ static void nvme_free_ns_head(struct kref *ref)
 	nvme_mpath_remove_disk(head);
 	ida_simple_remove(&head->subsys->ns_ida, head->instance);
 	list_del_init(&head->entry);
-	cleanup_srcu_struct(&head->srcu);
+	cleanup_srcu_struct_quiesced(&head->srcu);
 	kfree(head);
 }
 

  reply	other threads:[~2018-04-09 16:48 UTC|newest]

Thread overview: 19+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2018-04-02 12:37 [PATCH] nvme: fix flush dependency in delete controller flow Nitzan Carmi
2018-04-05  8:54 ` Christoph Hellwig
2018-04-05 13:14   ` Paul E. McKenney
2018-04-05 16:08     ` Paul E. McKenney
2018-04-05 19:55       ` Christoph Hellwig
2018-04-06  0:18         ` Paul E. McKenney
2018-04-06  6:19           ` Christoph Hellwig
2018-04-06 16:30             ` Paul E. McKenney
2018-04-08  7:20               ` Nitzan Carmi
2018-04-08 16:48                 ` Paul E. McKenney
2018-04-09  6:39                   ` Nitzan Carmi
2018-04-09  6:50                     ` Christoph Hellwig
2018-04-09 14:50                       ` [PATCH] nvme: avoid " Nitzan Carmi
2018-04-09 16:48                         ` Paul E. McKenney [this message]
2018-04-09 16:58                           ` Max Gurtovoy
2018-04-09 18:22                             ` Paul E. McKenney
2018-04-10 10:47                               ` Max Gurtovoy
2018-04-10 17:01                                 ` Paul E. McKenney
2018-04-09 16:30                     ` [PATCH] nvme: fix " Paul E. McKenney

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20180409164808.GK3948@linux.vnet.ibm.com \
    --to=paulmck@linux.vnet.ibm.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.