Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 8 Dec 2009 21:25:55 +0100
From:      Wiktor Niesiobedzki <bsd@w.evip.pl>
To:        Pawel Jakub Dawidek <pjd@freebsd.org>
Cc:        freebsd-geom <freebsd-geom@freebsd.org>
Subject:   Re: geli freezing 8.0 RELEASE and 7.2-STABLE
Message-ID:  <2ae8edf30912081225i224de5cg3a76beb3fda85ff7@mail.gmail.com>
In-Reply-To: <20091207213408.GE1795@garage.freebsd.pl>
References:  <2ae8edf30912050724i6f196e53y40ccd06970d59a7a@mail.gmail.com> <20091207213408.GE1795@garage.freebsd.pl>

next in thread | previous in thread | raw e-mail | index | archive | help
2009/12/7 Pawel Jakub Dawidek <pjd@freebsd.org>:
> Hi Wiktor:)
>
> My guess is that this is because GELI worker threads are running with
> too high priority (at least for your configuration).
>
> If I am right, the patch below should help you:
>
> =C2=A0 =C2=A0 =C2=A0 =C2=A0http://people.freebsd.org/~pjd/patches/g_eli.c=
.2.patch

Hi,

Yes, it looks like it is thread priority problem. With this patch
applied I do not see any freezes nor lags as before. I also did some
small check, how this affects the performance, and I see no
significant change.

I saw some similiar reports here:
http://forums.freebsd.org/showthread.php?t=3D6230 (though here it is
connected with ZFS and compression).

As far as I can imagine, what happens is, that ZFS usually sends data
to disks as bulk transfers (every 5-10 seconds), so when I'm using
GELI, it has quite a lot of data to encrypt, hence I observe some
lagging. But then, I'm not sure, if compression on ZFS could expose
this scenario in more aparent way.

After having a quick look on
http://www.freebsd.org/doc/en/books/arch-handbook/smp-design.html

Does it mean, that alawys, when we have a a PRIBIO kthread running
(geli thread encrypting data), lower priority processes (esp. userland
processes) will not get any CPU? Sounds bit different than what I
expected...


Cheers,

Wiktor Niesiobedzki



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?2ae8edf30912081225i224de5cg3a76beb3fda85ff7>