Only 86MB of RAM shown available on PS3 in Hardy

Bug #220524 reported by Dan Munckton on 2008-04-22
8
Affects Status Importance Assigned to Milestone
Ubuntu PS3 Port
Critical
Dan Munckton
Hardy
Critical
Unassigned
Intrepid
Critical
Dan Munckton
linux (Ubuntu)
Undecided
Dan Munckton
linux-ports (Ubuntu)
Undecided
Dan Munckton

Bug Description

Testing out the current hardy kernel on Playstation 3 I found that 'free -m' reports only 86MB of memory appears to be usable by the system. I noticed this while trying to work out why running X resulted in spontaneous reboots of X/GDM. My current theory is that it's hitting out of memory exceptions.

This is an HDD installation of Hardy and not just running from the LiveCD. The problem was not there in gutsy using the previous 'cell' specific kernel.

Related branches

Dan Munckton (munckfish) wrote :
Dan Munckton (munckfish) wrote :

Not attaching lspci log as it outputs nothing on ps3/hardy.

Dan Munckton (munckfish) wrote :
Dan Munckton (munckfish) wrote :
description: updated
Dan Munckton (munckfish) on 2008-04-22
Changed in ubuntu-ps3-port:
assignee: nobody → munckfish
status: New → Incomplete
Dan Munckton (munckfish) on 2008-04-22
Changed in ubuntu-ps3-port:
importance: Undecided → Critical
Dan Munckton (munckfish) wrote :

Running hardy under the old 'cell' kernel makes this problem and LP #220370 go away. Will now try to identify what it is we're missing in the new kernel.

Dan Munckton (munckfish) wrote :

Running with a custom build from the linux_2.6.24-16.30 sources using a .config generated by "make ps3_defconfig" this problem does not occur. Attaching dmesg output running this kernel build.

Dan Munckton (munckfish) wrote :
Dan Munckton (munckfish) wrote :

Attaching a comparison of the 2 kernel configs. Shows options that differ between them and options which are unique to one or the other. There are only a few notable differences AFAIKS. The majority of differences are where the ubuntu kernel configures code as modules rather than compiling in directly. The majority of options unique to the ubuntu config are the huge number of extra supported drivers. There are a handful of options which are unique only to ps3_defconfig.

From here I could now start testing each extra option in a rebuild of the ubuntu kernel to see which one solves the problem. If anyone out there has any ideas as to where best to focus energy please let me know. Now I have this report I will also try to contact upstream for ideas/input.

Cheers

Dan Munckton (munckfish) wrote :

I now have a patch from upstream [0] that I've tested and can confirm fixes this. Attached here for now. Awaiting details to access to kernel.ubuntu.com so we can setup a PS3/cell tree to push to for review.

The other option we have is to change the following config options. See [1] for explanation:

  CONFIG_SPARSEMEM_VMEMMAP=n
  CONFIG_FB_PS3_DEFAULT_SIZE_M=9
  CONFIG_PS3_USE_LPAR_ADDR=n

I've not tested these settings yet, and I'm still exploring what the side effects would be if any. May recommend to the kernel team that we move back to a ps3 specific custom kernel flavour.

[0] http://ozlabs.org/pipermail/cbe-oss-dev/2008-May/004704.html
[1] http://ozlabs.org/pipermail/cbe-oss-dev/2008-May/004721.html

Changed in ubuntu-ps3-port:
status: Incomplete → In Progress
Changed in linux:
assignee: nobody → munckfish
status: New → In Progress

The patch should do the trick without needing to use a customer flavour of the kernel.
I do recommend lowering the size of the PS3 FB though that can be done in a common
kernel.

I'm having paulus review the patch closely and send it to linus for .26. Once that's done
I'll look into backporting it to .25.x "stable" and .24.x "stable" but you can just stick it
in your package in the meantime.

Dan Munckton (munckfish) wrote :

Ben H's patch has been accepted into Linus' 2.6 tree.

http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=cec08e7a948326b01555be6311480aa08e637de2

The problem also affects our linux-ports kernel (based on 2.6.25.4). Although the effect is not quite as severe - only ~128 MB gets setup.

Can we pull this fix into linux-ports now? I don't have permission to do so.

Changed in linux-ports:
assignee: nobody → munckfish
status: New → In Progress

The Ubuntu Kernel Team is planning to move to the 2.6.27 kernel for the upcoming Intrepid Ibex 8.10 release. As a result, the kernel team would appreciate it if you could please test this newer 2.6.27 Ubuntu kernel. There are one of two ways you should be able to test:

1) If you are comfortable installing packages on your own, the linux-image-2.6.27-* package is currently available for you to install and test.

--or--

2) The upcoming Alpha5 for Intrepid Ibex 8.10 will contain this newer 2.6.27 Ubuntu kernel. Alpha5 is set to be released Thursday Sept 4. Please watch http://www.ubuntu.com/testing for Alpha5 to be announced. You should then be able to test via a LiveCD.

Please let us know immediately if this newer 2.6.27 kernel resolves the bug reported here or if the issue remains. More importantly, please open a new bug report for each new bug/regression introduced by the 2.6.27 kernel and tag the bug report with 'linux-2.6.27'. Also, please specifically note if the issue does or does not appear in the 2.6.26 kernel. Thanks again, we really appreicate your help and feedback.

On Fri, 2008-08-29 at 03:43 +0000, Leann Ogasawara wrote:
> The Ubuntu Kernel Team is planning to move to the 2.6.27 kernel for
> the
> upcoming Intrepid Ibex 8.10 release. As a result, the kernel team
> would
> appreciate it if you could please test this newer 2.6.27 Ubuntu
> kernel.

Leann thanks. But this affects the intrepid-ports kernel now, which is
stuck on 2.6.25 until some someone is motivated enough to upgrade it.

Jussi Saarinen (jms) wrote :

As I wasn't able to find package linux-image-2.6.27-* from (intrepid) repositories for powerpc I installed linux-source-2.6.27 and compiled kernel for PS3 using 'make ps3_defconfig' and 'make-kpkg --initrd kernel_image kernel_headers'.
Free reports 222028 KB of total memory.

Hi Dan,

Yup sorry about the noise here. Our automated call for testing hit a majority of the open kernel bugs - it was bound to happen the stock reply would not read well with some reports such as this one. I apologize for any confusion this may cause to the subscribers here. Thanks.

Dan Munckton (munckfish) wrote :

Patch sat in my Git tree awaiting journey to the official intrepid-ports tree.

Changed in linux-ports:
status: In Progress → Fix Committed
Launchpad Janitor (janitor) wrote :

This bug was fixed in the package linux-ports - 2.6.25-2.3

---------------
linux-ports (2.6.25-2.3) intrepid; urgency=low

  [Ben Collins]

  * build: Add 386 to d-i build

  [Benjamin Herrenschmidt]

  * PS3/powerpc vmemmap variable page size patch
    -LP: #220524

  [Dan Munckton]

  * build: Fix getabis script so it fetches the ports packages successfully
  * build: fix binary-headers target
  * config: enabled Gelic wireless support for PS3
    -LP: #194339
  * config: lowered the PS3 framebuffer size to 9MB
    -LP: #273710

  [Fabio M. Di Nitto]

  * Add GNBD driver
  * SAUCE: Export gfs2 symbols required for gfs1 kernel module
  * Add GFS driver
  * Enable GFS* config options

 -- Stefan Bader <email address hidden> Wed, 24 Sep 2008 14:50:53 -0400

Changed in linux-ports:
status: Fix Committed → Fix Released
Dan Munckton (munckfish) on 2008-10-09
Changed in linux:
status: In Progress → Invalid
To post a comment you must log in.
This report contains Public information  Edit
Everyone can see this information.

Other bug subscribers