2002-02-07 20:41:24

by Louis E Garcia II

[permalink] [raw]
Subject: Problem with rmap-12c


I tried rmap-12c and had lots of swap usage. I when back to 12a and
everything calmed down. Is their a known problem with 12c?

--Louis



2002-02-07 22:12:37

by Rik van Riel

[permalink] [raw]
Subject: Re: Problem with rmap-12c

On 7 Feb 2002, Louis Garcia wrote:

> I tried rmap-12c and had lots of swap usage. I when back to 12a and
> everything calmed down. Is their a known problem with 12c?

Nope, but the RSS limit enforcing stuff is a possible
suspect.

It turns out I used a "struct pte_t" in over_rss_limit(),
which turned into a compiler warning, for which I didn't
spot the cause ;)

A fix for the bug was sent by Roger Larsson, who spotted
the fact that "pte_t" already has a "struct" inside it.

Maybe page aging isn't working in rmap-12c because of this
stupid mistake ... but it's a long shot. Maybe I should
release rmap 12d tonight ? ;)

regards,

Rik
--
"Linux holds advantages over the single-vendor commercial OS"
-- Microsoft's "Competing with Linux" document

http://www.surriel.com/ http://distro.conectiva.com/

2002-02-07 22:31:59

by Shawn Starr

[permalink] [raw]
Subject: Re: Problem with rmap-12c

Please do :)

I've been noticing those same results.

Shawn.

On Thu, 2002-02-07 at 17:12, Rik van Riel wrote:
> On 7 Feb 2002, Louis Garcia wrote:
>
> > I tried rmap-12c and had lots of swap usage. I when back to 12a and
> > everything calmed down. Is their a known problem with 12c?
>
> Nope, but the RSS limit enforcing stuff is a possible
> suspect.
>
> It turns out I used a "struct pte_t" in over_rss_limit(),
> which turned into a compiler warning, for which I didn't
> spot the cause ;)
>
> A fix for the bug was sent by Roger Larsson, who spotted
> the fact that "pte_t" already has a "struct" inside it.
>
> Maybe page aging isn't working in rmap-12c because of this
> stupid mistake ... but it's a long shot. Maybe I should
> release rmap 12d tonight ? ;)
>
> regards,
>
> Rik
> --
> "Linux holds advantages over the single-vendor commercial OS"
> -- Microsoft's "Competing with Linux" document
>
> http://www.surriel.com/ http://distro.conectiva.com/
>
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at http://www.tux.org/lkml/
>
--
Shawn Starr
Developer Support Engineer
Datawire Communication Networks Inc.
10 Carlson Court, Suite 300
Toronto, ON, M9W 6L2
T: 416-213-2001 ext 179 F: 416-213-2008
8

2002-02-07 23:31:15

by Rik van Riel

[permalink] [raw]
Subject: Re: Problem with rmap-12c

On 7 Feb 2002, Shawn Starr wrote:
> On Thu, 2002-02-07 at 17:12, Rik van Riel wrote:
> > On 7 Feb 2002, Louis Garcia wrote:
> >
> > > I tried rmap-12c and had lots of swap usage. I when back to 12a and
> > > everything calmed down. Is their a known problem with 12c?
> >
> > Nope, but the RSS limit enforcing stuff is a possible
> > suspect.
> >
> > It turns out I used a "struct pte_t" in over_rss_limit(),
> > which turned into a compiler warning, for which I didn't
> > spot the cause ;)

> Please do :)
>
> I've been noticing those same results.

OK, uploaded. I'd love to hear if this stupid extra
'struct' statement was causing the trouble or if there's
a more fundamental problem with 12c.

cheers,

Rik
--
"Linux holds advantages over the single-vendor commercial OS"
-- Microsoft's "Competing with Linux" document

http://www.surriel.com/ http://distro.conectiva.com/

2002-02-08 21:06:14

by Louis E Garcia II

[permalink] [raw]
Subject: Re: Problem with rmap-12c

Ok, I've tried rmap-12d and the swaping is better but still worse then
12a. One thing I should say is I'm also using Andrews low latency patch.

Is you want vm stats let me know.

--Louis


On Thu, 2002-02-07 at 17:12, Rik van Riel wrote:
> On 7 Feb 2002, Louis Garcia wrote:
>
> > I tried rmap-12c and had lots of swap usage. I when back to 12a and
> > everything calmed down. Is their a known problem with 12c?
>
> Nope, but the RSS limit enforcing stuff is a possible
> suspect.
>
> It turns out I used a "struct pte_t" in over_rss_limit(),
> which turned into a compiler warning, for which I didn't
> spot the cause ;)
>
> A fix for the bug was sent by Roger Larsson, who spotted
> the fact that "pte_t" already has a "struct" inside it.
>
> Maybe page aging isn't working in rmap-12c because of this
> stupid mistake ... but it's a long shot. Maybe I should
> release rmap 12d tonight ? ;)
>
> regards,
>
> Rik
> --
> "Linux holds advantages over the single-vendor commercial OS"
> -- Microsoft's "Competing with Linux" document
>
> http://www.surriel.com/ http://distro.conectiva.com/
>


2002-02-08 21:36:36

by Rik van Riel

[permalink] [raw]
Subject: Re: Problem with rmap-12c

On 8 Feb 2002, Louis Garcia wrote:

> Ok, I've tried rmap-12d and the swaping is better but still worse then
> 12a. One thing I should say is I'm also using Andrews low latency patch.

OK, I'll try to find what is causing the regression in
performance. I'll try to mail you a patch later this
weekend.

regards,

Rik
--
Will hack the VM for food.

http://www.surriel.com/ http://distro.conectiva.com/

2002-02-16 21:29:08

by Louis E Garcia II

[permalink] [raw]
Subject: future of rmap VM

When do you plan to have a complete rmap VM ready for general use? Whats left to do?

Also, have you discussed merging this VM into the mainline kernel, either 2.4 or 2.5?

To me, this VM is far better than the current VM.

--Louis

2002-02-16 21:42:14

by Rik van Riel

[permalink] [raw]
Subject: Re: future of rmap VM

On 16 Feb 2002, Louis Garcia wrote:

> When do you plan to have a complete rmap VM ready for general use?
> Whats left to do?

Rmap version 11c is considered ready for use.

For rmap 12, which is also ready for use, I have added the
feature of RSS limit enforcement.

Rmap version 13 will get new page launder features.

It seems from now on -rmap will get new features, but no
more changes to the -rmap architecture are needed right
now.

> Also, have you discussed merging this VM into the mainline kernel,
> either 2.4 or 2.5?

I've started pushing some trivial stuff for inclusion into
2.5, but unfortunately current 2.5 just won't boot on my
test box so I cannot test it (either with or without the
patch).

Once 2.5 is working on my test box, I'll be ready to merge
some of the larger stuff ...

... I'll try to push things to Linus in small parts, so
nothing will break in the process.

> To me, this VM is far better than the current VM.

Good to hear I've got happy users.

cheers,

Rik
--
"Linux holds advantages over the single-vendor commercial OS"
-- Microsoft's "Competing with Linux" document

http://www.surriel.com/ http://distro.conectiva.com/