[Gta04-owner] Replicant GTA04 : suspend bug

[NeilBrown]

On Mon, 28 Jan 2013 18:18:24 +0400 Alexander Rumyantsev <ran at mail.wplus.net>
wrote:

> 
> > On Sun, 27 Jan 2013 22:23:39 +0100 Paul Kocialkowski <paulk at paulk.fr> wrote:
> > 
> > > Le lundi 28 janvier 2013 à 07:13 +1100, NeilBrown a écrit :
> > > > > Trying to investigate this problem I notice this patch
> > > > > https://github.com/radekp/linux-2.6/commit/1dcf9f5f6493e5278edb5eadb453c8b9c8ee6dc2
> > > > > After testing with Replicant-2.3 I found that I have not suspend/resume problem. So
> > > > > 1) patch may be a temporary solution
> > > > > 2) Radek or Nail - please, could you give more detail what was a reason you implement
> > > > > this patch and what was next steps.
> > > > 
> > > > The comment at the top nearly says it all:
> > > > 
> > > >     Don't switch consoles or sync on suspend.
> > > > 
> > > >     Neither of these actions are really needed and can slow suspend
> > > >     down, so disable them.
> > > > 
> > > > Applying the patch speed up suspend for me, and removed the switch to
> > > > text-console and back which was visually disturbing.  When you use
> > > > auto-suspend so that it happens quite often, this can become significant.
> > > > As there were no down-sides I just left it like that.
> > > > 
> > > > Did you find that reverted the patch fixed a problem?  Can you provide more
> > > > details of the problem.
> > > > This email seems like the middle of a conversation that I haven't seen the
> > > > rest of.  Did it take place on some other list that you can point me to?
> > > 
> > > We are still trying to find an usable Android kernel that would work
> > > with DM37x, and for now, all the combinations we've tried turned out to
> > > fail at some point, leaving us facing huge issues each time. We
> > > definitely want to add support for the GTA04 in Replicant, but we can't
> > > start any work on the userspace if we don't have a complete and usable
> > > kernel first. That's why there has been no announcement regarding GTA04
> > > even though we released Replicant 4.0 for other devices.
> > > 
> > > The context of that email was a conversation between Alexander and me
> > > regarding Replicant suspend issues with GTA04.
> > > 
> > > We'll keep trying to end up with a stable kernel that could be used with
> > > ICS, though that would be easier if there was a 3.x Android kernel for
> > > OMAP DM37x devices out there. GNUtoo is working to try new combinations
> > > while I gave up for now and focused on other devices.
> > > 
> > 
> > Suppose that - hypothetically - I wanted to spend a bit of time adding
> > Android bits to current mainline so that it would work on the GTA04 for
> > replicant, what would I look at adding.
> > 
> > i.e. do you have a list of what doesn't work if you try my latest kernel?
> > 
> > I'm guessing that suspend-blocker is the big one.  Much of the rest is
> > probably there, at least in 'staging'.
> > Do you have an fs image that shows up these missing bits clear so that if I
> > were - again, hypothetically - to try to enhance the kernel, I could easily
> > test if it were helping at all?
> > 
> > Thanks,
> > NeilBrown
> 
> Hello Neil
> 
> 1) Thank you for quick answer and explanation a patch above. Your explanation make me sure 
> in my understanding. This patch may be used as temporary solution of Replicant 2.3 suspend/resume 
> problem. But real problem is some deeper, console restored incorrectly sometime if this patch 
> not used.
> 2) My understanding a situation with Android/Replicant kernel is same as Paul has - we have not
> stable kernel. It is some kernel version, which recommended to different Android version:
> 2.6.35 - Gingerbread, 3.0.8 - ICS . So it is not good to follow latest version kernel 
> from Android point of view as you suggest. ( However Denis is going this way now ).
> It is not good that in current Replicant-2.3 is used 3.2.27, it is because nothing else was found.
> ( I can compare situation for example with RedHat Enterprise Linux - stable kernel for version ) 
> At moment I am trying to find which features from latest kernel versions is it possible to
> use with current used, Paul try go get working stage for 3.0.8 version. 
> My opinion that all correction concerning GTA04 should be available not only for latest kernel
> version, but for some stable release. I understand that it is difficult task, but we should discuss
> how to organize that.
> 
> Alexandre  
> 

Hi Alexandre,

 if you look at the 'hardware validation' kernel for the GTA04 you will find
 that it diverges very significantly from 2.6.32, on which it is based.
 About 1500 files changes and 100s of thousands of additions and deletions.
 This is what was required to get it to work on the OMAP SoC - and to get
 various other bits to work.  These presumably came from the OMAP development
 work, long before that work fully migrated to mainline.

 I had a quick look at the 3.0.8 Android kernel that Paul is working with
 and  it has significant elements of OMAP support missing.  In particular it
 does have (or at least doesn't enable) and CPU-IDLE support - the CPU module
 is always at full power.  I don't doubt there are many other issues which
 will make it impractical to create a really working phone.

 If you can find some vendor who sold an OMAP3 based phone running the
 relevant version of Android, then getting the kernel they used might be a
 reasonable starting point as they would presumably have added all the OMAP
 development code much like those in the 2.6.32 hardware-validation kernel.

 If not, the only real options are:
   1/ merge all the OMAP support from somewhere into an Android kernel
   2/ merge all the Android support into a mainline kernel.

 I strongly believe that the latter would be substantially simpler.  This is
 in part because the Android features were designed to be very modular, and
 so easy to move from on kernel to another, while the OMAP support is very
 deep and intertwined and really quite complex.

NeilBrown
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 828 bytes
Desc: not available
URL: <http://lists.goldelico.com/pipermail/gta04-owner/attachments/20130129/60346a2a/attachment-0001.bin>