SEO

Hi all,

I'm having a problem with HAL's `hald' process on an Intel Celeron 500
MHz fileserver running Slackware 12.0 that I'm managing.

The problem is that `hald' slowly keeps eating up more and more RAM over
time. Right after restarting it it'll consume 1.7% of the system's 192
MiB of RAM + 384 MiB of swap. Three days later it'll consume about 20%.
About a week after starting it, it'll use some 50%, meaning it has
allocated more RAM than the machine has.

Doing /etc/rc.d/rc.hald restart fixes it temporarily and makes it start
all over again. I guess I could have it restarted it from a daily
cronjob but that's fighting a symptom, not solving a problem.

Any ideas?

- Martijn

On Mon, 10 Dec 2007 09:22:39 +0100, Martijn Dekker wrote:

> Hi all,
>
> I'm having a problem with HAL's `hald' process on an Intel Celeron 500
> MHz fileserver running Slackware 12.0 that I'm managing.
>
> The problem is that `hald' slowly keeps eating up more and more RAM over
> time. Right after restarting it it'll consume 1.7% of the system's 192
> MiB of RAM + 384 MiB of swap. Three days later it'll consume about 20%.
> About a week after starting it, it'll use some 50%, meaning it has
> allocated more RAM than the machine has.
>
> Doing /etc/rc.d/rc.hald restart fixes it temporarily and makes it start
> all over again. I guess I could have it restarted it from a daily
> cronjob but that's fighting a symptom, not solving a problem.
>
> Any ideas?
>
> - Martijn
>
Since I have fixed my server (referenced in the post below), hald seems
to be working much better:
http://groups.google.com/group/alt.o...441a09505b3db8

--
Douglas Mayne

On 2007-12-10, Martijn Dekker <martijn@inlv.demon.nl> wrote:
> Hi all,
>
> I'm having a problem with HAL's `hald' process on an Intel Celeron 500
> MHz fileserver running Slackware 12.0 that I'm managing.
>
> The problem is that `hald' slowly keeps eating up more and more RAM over
> time. Right after restarting it it'll consume 1.7% of the system's 192
> MiB of RAM + 384 MiB of swap. Three days later it'll consume about 20%.
> About a week after starting it, it'll use some 50%, meaning it has
> allocated more RAM than the machine has.
>
> Doing /etc/rc.d/rc.hald restart fixes it temporarily and makes it start
> all over again. I guess I could have it restarted it from a daily
> cronjob but that's fighting a symptom, not solving a problem.


Well, if all this box is doing is functioning as a fileserver,
I would disable HAL entirely - it serves no purpose for you.

That being said, there were a lot of fixes in the hal release
that's now in -current, so you might try building that version
on your 12.0 box (be sure to build hal-info too - 0.5.10 requires
the newer hal-info vice versa) [1]

For the record, I've had a couple of boxes running extended periods
of time on stock 12.0 -- my desktop was up for somewhere around 80
days until last week -- and haven't noticed any issues. A detailed
bug report *might* be useful, especially if you can narrow down a
way to reproduce it reliably elsewhere, but the hal devs are going
to want that done on the latest stuff from svn, I'm sure.

[1] The idea with separating hal-info out from hal was that
distributions could upgrade the hal-info package and get
new most up to date device information files without having
to upgrade hal itself -- for distributions like RHEL and
such, this would be very nice -- however, it hasn't panned
out that way. Big surprise, I suppose

-RW

In article <Jjo7j.14245$Mu4.13532@bignews7.bellsouth.net>,
Robby Workman <newsgroups@rlworkman.net> wrote:

> On 2007-12-10, Martijn Dekker <martijn@inlv.demon.nl> wrote:
[...]
> > The problem is that `hald' slowly keeps eating up more and more RAM
> > over time.
[...]
> Well, if all this box is doing is functioning as a fileserver,
> I would disable HAL entirely - it serves no purpose for you.

Well, it would be nice if it were possible in a user-friendly way (i.e.
without manual mounting) for the owner to pop in a CD or DVD and copy
its contents to the server directly, without having to use the network.

> That being said, there were a lot of fixes in the hal release
> that's now in -current, so you might try building that version
> on your 12.0 box (be sure to build hal-info too - 0.5.10 requires
> the newer hal-info vice versa) [1]

Thanks, I've tried that. So far, so good -- in a few days to a week I'll
know whether it helped.

- Martijn

In article <pan.2007.12.10.13.30.17.755176@sl12.localnet>,
Douglas Mayne <doug@sl12.localnet> wrote:

> Since I have fixed my server (referenced in the post below), hald seems
> to be working much better:
> http://groups.google.com/group/alt.o...441a09505b3db8

Thanks, but since X does not normally run on this machine, that's
probably not relevant to this problem.

- Martijn

In article <martijn-3394C1.12533711122007@news.individual.net>,
Martijn Dekker <martijn@inlv.demon.nl> wrote:

> In article <Jjo7j.14245$Mu4.13532@bignews7.bellsouth.net>,
> Robby Workman <newsgroups@rlworkman.net> wrote:
[...]
> > That being said, there were a lot of fixes in the hal release
> > that's now in -current, so you might try building that version
> > on your 12.0 box (be sure to build hal-info too - 0.5.10 requires
> > the newer hal-info vice versa) [1]
>
> Thanks, I've tried that. So far, so good -- in a few days to a week I'll
> know whether it helped.

Nope, didn't help. After two days, hald from HAL 0.5.10 is already up to
8.6% of total memory space. It started out at 2.0%.

I'm stumped. AFAIK, I'm using the default configuration -- I didn't
change anything. Does anyone have any idea where to start looking?

- Martijn

On 2007-12-13, Martijn Dekker <martijn@inlv.demon.nl> wrote:
> In article <martijn-3394C1.12533711122007@news.individual.net>,
> Martijn Dekker <martijn@inlv.demon.nl> wrote:
>
>> In article <Jjo7j.14245$Mu4.13532@bignews7.bellsouth.net>,
>> Robby Workman <newsgroups@rlworkman.net> wrote:
> [...]
>> > That being said, there were a lot of fixes in the hal release
>> > that's now in -current, so you might try building that version
>> > on your 12.0 box (be sure to build hal-info too - 0.5.10 requires
>> > the newer hal-info vice versa) [1]
>>
>> Thanks, I've tried that. So far, so good -- in a few days to a week I'll
>> know whether it helped.
>
> Nope, didn't help. After two days, hald from HAL 0.5.10 is already up to
> 8.6% of total memory space. It started out at 2.0%.
>
> I'm stumped. AFAIK, I'm using the default configuration -- I didn't
> change anything. Does anyone have any idea where to start looking?


Well, crap.

I've never used valgrind, so I honestly have no idea how to go about
employing it, but supposedly this idea is its purpose:
http://slackbuilds.org/repository/12...ment/valgrind/

This is probably worth a mail to the hal mailing list:
http://lists.freedesktop.org/mailman/listinfo/hal
to see if the devs know of any existing leaks and get some other
troubleshooting ideas.

-RW