Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 30 Dec 2004 11:45:11 -0700
From:      Scott Long <scottl@freebsd.org>
To:        Bosko Milekic <bmilekic@technokratis.com>
Cc:        "Christian R." <cr@it.dk>
Subject:   Re: Memory handling problem with FreeBSD 5.3?
Message-ID:  <41D44CB7.8050805@freebsd.org>
In-Reply-To: <20041230181118.GA71221@technokratis.com>
References:  <2rg6t0ll6th99497btiu13foetanfrn0gr@4ax.com> <41D368A4.1030408@freebsd.org> <8tg7t09g1f04igli4cplkob6ens98ckou0@4ax.com> <aqn7t0913lio6dv4nrbj0f7f73uc8kvsr0@4ax.com> <qh98t0he9573makq9l542tpkmpbujot01n@4ax.com> <41D43DC7.8070502@freebsd.org> <20041230181118.GA71221@technokratis.com>

index | next in thread | previous in thread | raw e-mail

Bosko Milekic wrote:
> On Thu, Dec 30, 2004 at 10:41:27AM -0700, Scott Long wrote:
> 
>>Christian R. wrote:
>>
>>>On Thu, 30 Dec 2004 12:02:16 +0100, you wrote:
>>>
>>>
>>>
>>>>Would the AMD64 version of FreeBSD 5.3 be a better choice?
>>>
>>>
>>>I have now installed the AMD64 version on the server. It has been
>>>running stable now for some hours with alot of load and all the 6 GB
>>>activated. It seems to be a problem with the i386 version og FreeBSD
>>>5.3.
>>>
>>
>>i386 and amd64 have the exact same code for handling the needs of the
>>amr driver.  I'm a bit worried that one works for you while the other
>>doesn't; I extensively tested the code on both platforms with 8GB of
>>RAM.
>>
>>Scott
> 
> 
>   Maybe I missed something in this thread, but what does it have to do
>   with the amr driver?  The first thing that came to mind was something
>   with PAE and how it affects PDE modifications.
> 
>   On i386, it might be worth trying these combinations:
> 
>   1. SMP, NO PAE.
>   2. UP, PAE.
>   3. UP, NO PAE.
>   
>   As we already know that "SMP, PAE" for you has problems.
> 

Bounce buffers were horribly broken in 5.3-RELEASE for amd64 and i386,
and the amr driver itself was unreliable in using them even when they
worked (the amr driver cannot do 64-bit S/G right now).  I fixed all of
this stuff a few weeks ago.  I tested all of this extensively under PAE,
but you're correct that this particular problem could very well be
related to pmap problems specific to i386.

Scott


home | help

Want to link to this message? Use this
URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?41D44CB7.8050805>