linux-nvme.lists.infradead.org archive mirror
 help / color / mirror / Atom feed
From: hch@infradead.org (Christoph Hellwig)
Subject: [PATCHv2 1/3] nvme: Remove RCU namespace protection
Date: Tue, 28 Jun 2016 01:31:37 -0700	[thread overview]
Message-ID: <20160628083137.GA32618@infradead.org> (raw)
In-Reply-To: <1466702946-13065-2-git-send-email-keith.busch@intel.com>

> +static struct nvme_ns *nvme_get_ns(struct nvme_ctrl *ctrl, unsigned nsid)

Can you call this nvme_find_get_ns?  A plain get is usually just
a wrapper around kref_get.

>   done:
> +	mutex_lock(&ctrl->namespaces_mutex);
>  	list_sort(NULL, &ctrl->namespaces, ns_cmp);
>  	mutex_unlock(&ctrl->namespaces_mutex);

This now leaves the list unordered between lock drops (actually we
already had that issue with the RCU conversion).  I think we just need
to do an already sorted insert, which shouldn't be too hard.

> @@ -1656,10 +1662,8 @@ void nvme_remove_namespaces(struct nvme_ctrl *ctrl)
>  	if (ctrl->state == NVME_CTRL_DEAD)
>  		nvme_kill_queues(ctrl);
>  
> -	mutex_lock(&ctrl->namespaces_mutex);
>  	list_for_each_entry_safe(ns, next, &ctrl->namespaces, list)
>  		nvme_ns_remove(ns);
> -	mutex_unlock(&ctrl->namespaces_mutex);

And this is the scary one - it does an unprotected
list_for_each_entry_safe, and nvme_remove_namespaces isn't even called
from the scan workqueue.

I think this needs to be something like:

	mutex_lock(&ctrl->namespaces_mutex);
	list_splice_init(&ctrl->namespaces, &tmp);
	mutex_unlock(&ctrl->namespaces_mutex);

	list_for_each_entry_safe(ns, next, &tmp, list) {
		..

		nvme_ns_remove(ns);

> +	mutex_lock(&ctrl->namespaces_mutex);
> +	list_for_each_entry(ns, &ctrl->namespaces, list) {
>  		if (!kref_get_unless_zero(&ns->kref))
>  			continue;

> @@ -1848,7 +1852,7 @@ void nvme_kill_queues(struct nvme_ctrl *ctrl)
>  
>  		nvme_put_ns(ns);

The get/put pair here can go away now as there will always be at
least one reference to the ns if it is on the ->namespaces list.

  reply	other threads:[~2016-06-28  8:31 UTC|newest]

Thread overview: 12+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-06-23 17:29 [PATCH 0/3] Namespace iteration fixes Keith Busch
2016-06-23 17:29 ` [PATCHv2 1/3] nvme: Remove RCU namespace protection Keith Busch
2016-06-28  8:31   ` Christoph Hellwig [this message]
2016-06-28 16:35     ` Keith Busch
2016-06-30  6:48       ` Christoph Hellwig
2016-06-30 14:57         ` Keith Busch
2016-06-30 22:59           ` Keith Busch
2016-06-23 17:29 ` [PATCH 2/3] nvme: Kill detached namespaces prior to removal Keith Busch
2016-06-28  8:32   ` Christoph Hellwig
2016-06-23 17:29 ` [PATCH 3/3] nvme: Put invalid namespaces on removal list Keith Busch
2016-06-28  8:32   ` Christoph Hellwig
2016-06-23 17:44 ` [PATCH 0/3] Namespace iteration fixes Keith Busch

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=20160628083137.GA32618@infradead.org \
    --to=hch@infradead.org \
    /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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).