linuxppc-dev.lists.ozlabs.org archive mirror
 help / color / mirror / Atom feed
From: Marco Stornelli <marco.stornelli@coritel.it>
To: Ruksen INANIR <rukseninanir@gmail.com>
Cc: linuxppc-embedded@ozlabs.org
Subject: Re: increase lowmem value for ppc
Date: Thu, 17 Jul 2008 14:05:56 +0200	[thread overview]
Message-ID: <487F35A4.7050203@coritel.it> (raw)
In-Reply-To: <487F3248.9080405@gmail.com>

Ruksen INANIR ha scritto:
> 
> Thanks for your comments.
> I tried the highmem option but one of the device drivers does not 
> support highmem, so it did not work.

Strange....you can try to look at it, it could be only some GFP flag.

> the only choice for me is to increase the lowmem to 1520 MB. the maximum 
> lowmem value i reached succesfully is 1503 MB. but any value more than 
> 1503 results kernel not able to load kernel modules.

Mmmm....have you tried a stress test? It sounds like a bit strange that 
the kernel works well with this mapping but without other modification.

> Upgrading to a 2.6.x kernel may solve the problem, but it needs a lot of 
> effort. Does 2.6.x kernel support 2G physical memory?
> 
It's the same. Some architecture have 36-bit addressing, but I don't 
know if the latest kernel has the support for this feature for powerpc.
> 
> 
> 
> Marco Stornelli wrote:
>> Ruksen INANIR ha scritto:
>>>
>>> Is there any side effects of changing the address splitting? 
>>
>> Yes, in this way the application address space is smaller than normal. 
>> Some big applications (for example same dbms) might not work. Usually 
>> with 2G/2G most of the applications work well.
>>
>>> What should i change for 2G/2G splitting? should i shift the kernel 
>>> start?
>>> Thanks
>>>
>> Yes in advanced options of the kernel menu, but sometimes it's not 
>> easy.  Sometimes ago I had to do the same thing but I have to change 
>> the kernel code because some operations were hard-coded. Be careful 
>> because it's not an easy operation. However I'd suggest you to use the 
>> highmem, the memory performance are worse than direct memory mapping, 
>> but usually it's not a problem because the overhead is low.
>>
>>>
>>> Marco Stornelli wrote:
>>>> Marco Stornelli ha scritto:
>>>>> Ruksen INANIR ha scritto:
>>>>>>
>>>>>> Is there a way to increase the lowmem value for ppc. The 
>>>>>> MAX_LOW_MEM is defined as max 768 MB. But a value around 1.5 G 
>>>>>> works with no problem. But when i try to increase this value to 
>>>>>> 1520 or more, kernel complaints about no space for memory 
>>>>>> allocation when loading kernel modules.
>>>>>> I do not want to use HIGHMEM config. What is the max lowmem value 
>>>>>> for a ppc system? What other setting are needed to use 1520 MB (or 
>>>>>> more) as lowmem .
>>>>>> The ppc card has 2G on board memory. 2.4.22 kernel is used.
>>>>>>
>>>>>> Thanks
>>>>>> _______________________________________________
>>>>>> Linuxppc-embedded mailing list
>>>>>> Linuxppc-embedded@ozlabs.org
>>>>>> https://ozlabs.org/mailman/listinfo/linuxppc-embedded
>>>>>>
>>>>> With 32-bit arch you may not use more than 1GB to map the memory 
>>>>> (minus some space for some kernel operation). The value of 768MB 
>>>>> was not there by chance.
>>>>>
>>>> Only an additional comment: I meant with the address splitting 
>>>> 3G/1G. To map more than 1GB, you have to change the splitting 2G/2G 
>>>> for example.
>>>>
>>>
>>> _______________________________________________
>>> Linuxppc-embedded mailing list
>>> Linuxppc-embedded@ozlabs.org
>>> https://ozlabs.org/mailman/listinfo/linuxppc-embedded
>>>
>>
>>
> 
> _______________________________________________
> Linuxppc-embedded mailing list
> Linuxppc-embedded@ozlabs.org
> https://ozlabs.org/mailman/listinfo/linuxppc-embedded
> 


-- 
Marco Stornelli
Embedded Software Engineer
CoRiTeL - Consorzio di Ricerca sulle Telecomunicazioni
http://www.coritel.it

marco.stornelli@coritel.it
+39 06 72582838

  reply	other threads:[~2008-07-17 12:03 UTC|newest]

Thread overview: 8+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2008-07-17 10:32 increase lowmem value for ppc Ruksen INANIR
2008-07-17 11:14 ` Marco Stornelli
2008-07-17 11:21   ` Marco Stornelli
2008-07-17 11:22     ` Ruksen INANIR
2008-07-17 11:35       ` Marco Stornelli
2008-07-17 11:51         ` Ruksen INANIR
2008-07-17 12:05           ` Marco Stornelli [this message]
     [not found] <mailman.2286.1216296226.2883.linuxppc-embedded@ozlabs.org>
2008-07-17 20:07 ` Siva Prasad

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=487F35A4.7050203@coritel.it \
    --to=marco.stornelli@coritel.it \
    --cc=linuxppc-embedded@ozlabs.org \
    --cc=rukseninanir@gmail.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;
as well as URLs for NNTP newsgroup(s).