Distributed Replicated Block Device (DRBD) development
 help / color / mirror / Atom feed
From: Philipp Reisner <philipp.reisner@linbit.com>
To: drbd-dev@lists.linbit.com
Subject: Re: [Drbd-dev] [DRBD 8.0 PATCH] Update state processing so that after-state-change is always done in worker thread
Date: Tue, 15 Jan 2008 10:14:07 +0100	[thread overview]
Message-ID: <200801151014.07561.philipp.reisner@linbit.com> (raw)
In-Reply-To: <342BAC0A5467384983B586A6B0B3767107C5AF5E@EXNA.corp.stratus.com>

Am Freitag, 11. Januar 2008 16:22:15 schrieb Graham, Simon:
> This patch updates the state processing so that the after state change
> processing is always done on the worker thread - my main motivation for
> this was to ensure that state change notifications are never re-ordered.
>
>
> This also involves the following:
> 1. starting the worker thread is done inline in drbd_set_state
> 2. the worker will be started whenever it is needed rather than
>    only when certain states are reached.
> 3. Marking the meta data dirty is done inline in drbd_set_state
>
> Simon

Hi Simon,

I definitely want to pick up that one, here are my notes from my
first review:

> @@ -818,19 +830,18 @@ int _drbd_set_state(drbd_dev* mdev, drbd_state_t
> ns,enum chg_state_flags flags) os.peer == Secondary && ns.peer == Primary)
>  		set_bit(CONSIDER_RESYNC, &mdev->flags);
>
> -	if( flags & ScheduleAfter ) {
> -		struct after_state_chg_work* ascw;
> -
> -		ascw = kmalloc(sizeof(*ascw), GFP_ATOMIC);
> -		if(ascw) {
> -			ascw->os = os;
> -			ascw->ns = ns;
> -			ascw->flags = flags;
> -			ascw->w.cb = w_after_state_ch;
> -			drbd_queue_work(&mdev->data.work,&ascw->w);
> -		} else {
> -			WARN("Could not kmalloc an ascw\n");
> -		}
> +	/* Make sure worker thread is running -- this is a NOP if it is already running */ 
> +	drbd_thread_start(&mdev->worker); 
> +

It is not allowed to call drbd_thread_start() while we are in _drbd_set_state(),
because drbd_thread_start() uses a sleeping function (wait_for_completion()).
And _drbd_set_state() is running with the req_lock hold.

But I think that can be easily took out of _drbd_set_state(), and done in the
context of the netlink connector callbacks.

I will prepare that...

-Phil
-- 
: Dipl-Ing Philipp Reisner                      Tel +43-1-8178292-50 :
: LINBIT Information Technologies GmbH          Fax +43-1-8178292-82 :
: Vivenotgasse 48, 1120 Vienna, Austria        http://www.linbit.com :

  parent reply	other threads:[~2008-01-15  9:14 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2008-01-11 15:22 [Drbd-dev] [DRBD 8.0 PATCH] Update state processing so that after-state-change is always done in worker thread Graham, Simon
2008-01-14  9:05 ` Lars Ellenberg
2008-01-15  9:14 ` Philipp Reisner [this message]
2008-01-15 13:50 ` Graham, Simon

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=200801151014.07561.philipp.reisner@linbit.com \
    --to=philipp.reisner@linbit.com \
    --cc=drbd-dev@lists.linbit.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