[ROM] Olympus CyanogenMOD 9 (with Kernel 3.1.10)

[PolesApart]

Unless @krystianp can debug the auto cpu hotplug feature it's probably not as stable as I though: it's very likely to be the main responsible for that 1 ~ 2 reboots a day (or more for some folks). It's been a while since I've seen an almost 48h uptime (with several phone usage scenarios: multimedia dock & wifi, car dock & waze & 3g, battery powered & 3g, battery powered & wifi), but since I disabled auto cpu hotplug and recompiled (I also disabled CONFIG_SWAP and accidentally set CONFIG_NVMAP_HIGHMEM_ONLY=y, but it's probably unrelated):

[email protected]:/data/data/berserker.android.apps.sshdroid/home # uptime
13:57:16 up 1 day, 22:59, load average: 0.16, 0.26, 0.40
[email protected]:/data/data/berserker.android.apps.sshdroid/home #

(the load is so high probably because it's docked right now)

I'll keep it running for a few more days and will then try without CONFIG_NVMAP_HIGHMEM_ONLY to see if it makes any difference. Too bad my other atrix is unusable, as I could test it right now.

 

[krystianp]

...
with the kernel kernel-2015_03_19_08_07_cm9.zip, I get the following issue check the picture.

is it only with that kernel? Is happening regularily?

...It comes when the phone is long time in deepsleep and sometimes by wake-up it reboots (probably the cpu-hot-plug). I get about 3 to 4 reboot every day with this kernel.

So those reboots happen only when waking up or also during normal/heavy usage?

I'll keep it running for a few more days and will then try without CONFIG_NVMAP_HIGHMEM_ONLY to see if it makes any difference. Too bad my other atrix is unusable, as I could test it right now.

Is it still running without reboots?

P.S. I officially killed my jtag'ed Atrix yesterday I'll make new one from my spare semi working one. It doesn't have audio working but for early kernel logs it's fine.

 

[PolesApart]

Is it still running without reboots?

Nope. It rebooted this dawn. I've got a last_kmsg capture but looking at it I don't yet have a clue where the problem could be. It ends barking about missing sdhci interrupts, but before that it has crazy OOM killer activity, wifi barking, binder complaining and all sorts of things. I'm attaching in case you want to take a look. last_kmsg tends not to be highly useful but I don't have a jtag or anything.

I'll start testing the same version without CONFIG_NVMAP_HIGHMEM_ONLY today, as if that has no impact on stability, it should allow more 3d apps to run (not that I noticed anything out of ordinary, but I just play dumb games now and then).

 

[krystianp]

@PolesApart Do you have sdcard inserted? If so, have you excluded possibility that it's because of that?

 

[PolesApart]

@PolesApart Do you have sdcard inserted? If so, have you excluded possibility that it's because of that?

I have a sdcard, and something like it's firmware halting could be the culprit indeed (it could be the wifi too, it's sdhci right?). If that's the case, too bad it leads to a reboot, normally drivers can cope with that (by resetting the sdhci controller and vanishing the block device, followed by it's reappearing, fsck & remount). But if it's just that it should be seldom, and if it's the card and not the wifi, cardless folks are going to enjoy very long uptimes.

I'll keep an eye on the logs for the next days. Also, I think I'll soon be able to revive my other atrix, I can leave it on for tests.

 

[todobarato]

Hello krystianp,

is it only with that kernel? Is happening regularily?
Yes.

So those reboots happen only when waking up or also during normal/heavy usage?
It was mostly by wake-up

But I think I found the reason, after disable the KSM in performance menu, the issue looks like is solved, today I have no reboots. And if KSM is enable it doest go into deepsleep

I will test longer to be sure that it works.

Sent from my MB860 using XDA Free mobile app

 

[krystianp]

Hello krystianp,

is it only with that kernel? Is happening regularily?
Yes.

So those reboots happen only when waking up or also during normal/heavy usage?
It was mostly by wake-up

But I think I found the reason, after disable the KSM in performance menu, the issue looks like is solved, today I have no reboots. And if KSM is enable it doest go into deepsleep

I will test longer to be sure that it works.

After you do that you can try new kernel. It's on my google drive i mentioned before. New kernel should work on cm9, atrics and hopefully on kk. It has different kind of potential solution to dsi issue and new version of cpu hotplug. And yes, I'm working also on atrix audio problems and yes I'm working on razr too

I've also noticed that how well cpu hotplug works depends on governor. Best option is conservative, at least for 456MHz for cpu off and 812MHz for cpu on. It might be necessary to adjust those margins differently for each governor.

 

[palmbeach05]

Hello krystianp,


After you do that you can try new kernel. It's on my google drive i mentioned before. New kernel should work on cm9, atrics and hopefully on kk. It has different kind of potential solution to dsi issue and new version of cpu hotplug. And yes, I'm working also on atrix audio problems and yes I'm working on razr too

I thought i saw your name over there at some point. Lol. Are you on the 926 or 925? Jw

Sent from my ATRIX HD using XDA Free mobile app

 

[krystianp]

I thought i saw your name over there at some point. Lol. Are you on the 926 or 925? Jw

Sent from my ATRIX HD using XDA Free mobile app

I went through all First had xt926M then bricked it, replaced board with xt926 but it came damaged (wifi/gps antenna on was damaged during transportation) then i switched to xt925 that I'm currently using as daily).

 

[bobmorsy]

@krystianp I can't get past the boot logo with the 26/3 kk kernel. I tried several times and waited for like 15 mins.
any clue?

Sent from my MB860 using XDA Free mobile app

 

[palmbeach05]

I went through all First had xt926M then bricked it, replaced board with xt926 but it came damaged (wifi/gps antenna on was damaged during transportation) then i switched to xt925 that I'm currently using as daily).

Oh ok lol, not bad. Sucks that that happened with the antenna, but eh, stuff happens. I know its a bit off topic, but how is the 3.0 kernel compared to the 3.4 kernel?

 

[krystianp]

I can't get past the boot logo with the 26/3 kk kernel. I tried several times and waited for like 15 mins.
any clue?

I was affraid of that. I used my @todobarato config with only small change for wifi. I suspected kk needs some more changes in config. I'll rebuild it this evening and upload it again.

Oh ok lol, not bad. Sucks that that happened with the antenna, but eh, stuff happens. I know its a bit off topic, but how is the 3.0 kernel compared to the 3.4 kernel?

From user perspective it doesn't really differ. I'm sure there are many improvements but can't really say much about it. I've just recently started looking at 3.4 kernel sources.

 

[PolesApart]

I was affraid of that. I used my @todobarato config with only small change for wifi. I suspected kk needs some more changes in config. I'll rebuild it this evening and upload it again.



From user perspective it doesn't really differ. I'm sure there are many improvements but can't really say much about it. I've just recently started looking at 3.4 kernel sources.

From my (little) experience with 3.4, it behaves way better than 3.0/3.1 at flushing dirty pages (it's more smooth, 3.0 was a regression compared to 2.6) on both foreground and background cases, and also in dealing with low memory (where it needs to flush dirty data and also search for pages to discard).

Recent phones behaves way better on both accounts (but most of the ones I played had 3.10 kernel), but I suspect that they have recent and (way) faster eMMC, and it may also play it's part on the equation.

Is there any 3.4 kernel for android with (already working) tegra2 drivers?

 

[nozorrog]

From my (little) experience with 3.4, it behaves way better than 3.0/3.1 at flushing dirty pages (it's more smooth, 3.0 was a regression compared to 2.6) on both foreground and background cases, and also in dealing with low memory (where it needs to flush dirty data and also search for pages to discard).

Recent phones behaves way better on both accounts (but most of the ones I played had 3.10 kernel), but I suspect that they have recent and (way) faster eMMC, and it may also play it's part on the equation.

Is there any 3.4 kernel for android with (already working) tegra2 drivers?

We have some 3.8 kernel for the Acer Iconia Tab a500 based on the same hardware as our Atrix 4G. Isn't possible to build a proper kernel for us? Thanx for advance.
http://forum.tegraowners.com/viewforum.php?f=9&sid=38a4b0f4dfc429c69e65bd76d9ae4f2d

 

[PolesApart]

We have some 3.8 kernel for the Acer Iconia Tab a500 based on the same hardware as our Atrix 4G. Isn't possible to build a proper kernel for us? Thanx for advance.
http://forum.tegraowners.com/viewforum.php?f=9&sid=38a4b0f4dfc429c69e65bd76d9ae4f2d

First of all: is it stable? If not, it's not even an option.
Assuming it is stable, then it's hard to say if it's feasible. If we are so lucky that all needed drivers are present and flatened-device-tree-aware (CONFIG_USE_OF & related options), it would be just a matter of setting some text files for binding devices to resources and voila.

Real life tends to be tricker than that. Think the motorola xoom: it's probably the most similar hardware to atrix (even considering it has a great deal of differences) and AFAIK they still run it on 2.6 kernel, and it's not for nothing.

So I would expect at least a bit of hardcore porting and debugging required. With just a little of that, we're probably talking a lot of work, but would still be feasible.

Now, if it's a little cruder than that, then huge amounts of porting / merging / debugging is required. I can only speak for myself, of course, but I don't think that anybody else here have that amount of free time in their hands...

The thing is that it's not a "try and succeed" task, meaning no guarantees it would result in a overall better experience (sometimes for several reasons the newer kernel performs poorer and it's hard to fix/optimize). Also, the kernel <-> userspace layer could have incompatibilities. Linux APIs changes a lot. Things like sound on newer kernel could require library changes. Some of these libraries aren't exactly open source and sometimes there's a way around, but sometimes not. I'm not sure if that's the case or not.

I'm afraid this is getting too off topic, can we ressurrect the kernel thread or something should anyone is really interested in this endeavor?

 

[palmbeach05]

It's hard to say. If we are so lucky that all needed drivers are present and flatened-device-tree-aware (CONFIG_USE_OF & related options), it would be just a matter of setting some text files for binding devices to resources and voila.

Real life tends to be tricker than that. Think the motorola xoom: it's probably the most similar hardware to atrix (even considering it has a great deal of differences) and AFAIK they still run it on 2.6 kernel, and it's not for nothing.

So I would expect at least a bit of hardcore porting and debugging required. With just a little of that, we're probably talking a lot of work, but would still be feasible.

Now, if it's a little cruder than that, then huge amounts of porting / merging / debugging is required. I can only speak for myself, of course, but I don't think that anybody else here have that amount of free time in their hands...

The thing is that it's not a "try and succeed" task, meaning no guarantees it would result in a overall better experience (sometimes for several reasons the newer kernel performs poorer and it's hard to fix/optimize). Also, the kernel <-> userspace layer could have incompatibilities. Linux APIs changes a lot. Things like sound on newer kernel could require library changes. Some of these libraries aren't exactly open source and sometimes there's a way around, but sometimes not. I'm not sure if that's the case or not.

I'm afraid this is getting too off topic, can we ressurrect the kernel thread or something should anyone is really interested in this endeavor?

My bad for that, i was looking for a quick opinion from krystian about the difference between kernels on our devices (we both have devices from the same chipset family). It then spilled into a discussion about upgrading our kernel here on the A4G.

It is an interesting theory though. I do have to agree I have seen better performance out of newer kernel (version number i mean). While each release on our 3.1.10 does get better and better, I do wonder if we should look into the possibility of updating it to a higher version number. Could it improve our device? Yes, it could. But it could also lead to issues, esp with as you were saying, libraries that aren't open sourced. There's also the age of the technology to consider as well. Will things break with an increased revision? I fully back an improved kernel, esp based off of our device's past track record to get to this point.

Sent from my ATRIX HD using XDA Free mobile app

 

[krystianp]

I have somewhere partially ported 3.4 kernel. I used tegra4 kernel from nvidia git server. I stopped at some point, afair most of olympus code was implemented but kernel had some compilation errors related to tegra, some parts weren't resolved for t2. Maybe it was updated since then, i can upload it on github if anyone wants to look at it.

I've uploaded kk kernel, should work now.

 

[vicfuze]

With today's kernel, kernel-2015_03_26_06_56_atrics.zip

• Wifi is working
• USB storage now works for me
• Camera still zippy
• Now testing battery life

 

[bobmorsy]

@krystianp I still can't get past the boot logo using the latest kk kernel.
I wiped cache and dalvic then flashed the zip in cwm over kk v4...am I doing something wrong?

Sent from my MB860 using XDA Free mobile app

 

[krystianp]

I still can't get past the boot logo using the latest kk kernel.
I wiped cache and dalvic then flashed the zip in cwm over kk v4...am I doing something wrong?

I suspect I'm still not using correct config. I'll check config from that first build for kk and use that.

Also I'd like to mention that there is a bug somewhere in kernel. Somewhere in code there is a cap on CPU with 760MHz value. Unfortunately it's not removed so CPU is stuck at 760MHz max. That might lead to thinking that CPU hotplug is failing to boot second core. But this time it isn't, hotplug boots second core @812MHz which is never reached. I'll locate that 760MHz cap and find why it's not removing it when needed.