linux-raid.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: NeilBrown <neilb@suse.com>
To: Zhilong Liu <zlliu@suse.com>, Shaohua Li <shli@kernel.org>
Cc: shli@fb.com, linux-raid@vger.kernel.org,
	Guoqing Jiang <gqjiang@suse.com>
Subject: Re: [PATCH] md:array cannot be opened again after 'md_set_readonly'
Date: Wed, 29 Mar 2017 08:18:02 +1100	[thread overview]
Message-ID: <87bmsl6qyd.fsf@notabene.neil.brown.name> (raw)
In-Reply-To: <0c5289f3-0160-8d2d-a5d8-da126b2fb468@suse.com>

[-- Attachment #1: Type: text/plain, Size: 1930 bytes --]

On Tue, Mar 28 2017, Zhilong Liu wrote:

> On 03/28/2017 02:22 AM, Shaohua Li wrote:
>> On Mon, Mar 27, 2017 at 03:52:25PM +0800, Zhilong Liu wrote:
>>> This is a bug about array cannot be opened again after 'md_set_readonly',
>>> because the MD_CLOSING bit is still waiting for clear.
>>> MD_CLOSING should only be set for a short period or time to avoid certain
>>> races. After the operation that set it completes, it should be cleared.
>> where is the bit set? Why don't clear it after the operation but clear it in
>> set_readonly?
>>   
>
> it goes two paths after set MD_CLOSING bit, do_md_stop and md_set_readonly,
> the do_md_stop has done the md_clean to clear the mddev->flags, thus I 
> put it in
> md_set_readonly.
> Thanks for your suggestion, is the following changing good for you? here 
> it has
> finished what it needs to do, so clear the MD_CLOSING bit in time.
> if looks good, I would send this in new revision patch.
>
> diff --git a/drivers/md/md.c b/drivers/md/md.c
> index f6ae1d6..e8c1130 100644
> --- a/drivers/md/md.c
> +++ b/drivers/md/md.c
> @@ -6868,6 +6868,7 @@ static int md_ioctl(struct block_device *bdev, 
> fmode_t mode,
>                  set_bit(MD_CLOSING, &mddev->flags);
>                  mutex_unlock(&mddev->open_mutex);
>                  sync_blockdev(bdev);
> +               clear_bit(MD_CLOSING, &mddev->flags);

No.

MD_CLOSING is used to prevent a race with md_open().
md_open() blocks on open_mutex(), but we cannot hold that here for
complicated reasons.
So we set MD_CLOSING and need to keep it set at least until
md_set_readonly() or do_md_stop() take that lock again.

Clearing it here just opens up the race again.

It needs to be cleared in md_ioctl, after any call to md_set_readonly()
or do_md_stop().
I've already posted a sample patch which has been tested.  Please don't
do something completely different.

NeilBrown

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 832 bytes --]

  parent reply	other threads:[~2017-03-28 21:18 UTC|newest]

Thread overview: 6+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2017-03-27  7:52 [PATCH] md:array cannot be opened again after 'md_set_readonly' Zhilong Liu
2017-03-27 18:22 ` Shaohua Li
2017-03-28 13:19   ` Zhilong Liu
2017-03-28 16:12     ` Shaohua Li
2017-03-28 21:18     ` NeilBrown [this message]
2017-03-28 21:13 ` NeilBrown

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=87bmsl6qyd.fsf@notabene.neil.brown.name \
    --to=neilb@suse.com \
    --cc=gqjiang@suse.com \
    --cc=linux-raid@vger.kernel.org \
    --cc=shli@fb.com \
    --cc=shli@kernel.org \
    --cc=zlliu@suse.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 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).