Re: [Announce] kernel-plus for Harmattan
 

Show us ;)

Re: [Announce] kernel-plus for Harmattan
 

haha, that's what I call a debug console xD

Re: [Announce] kernel-plus for Harmattan
 

Hello! Do i need input modprobe wl12xx_sdio_test for working inject mode?

Re: [Announce] kernel-plus for Harmattan
 

no, kernel is compiled to load needed modules

Re: [Announce] kernel-plus for Harmattan
 

run to see loaded modules ;)

btw. nfs server support modules are not loaded on startup...

thanks to nieldk I have some binaries to play with but so far I didn't have the time to set it up.
If that works maybe the kernel could be compiled to load the modules automatically....

EDIT:

ah, so that's that.... autoload on demand?

Re: [Announce] kernel-plus for Harmattan
 

Update:

Found the commits needed to introduce multiple-VIF support into wl12xx driver. (in response to this post)

I'll check whether mac80211 as present in current 2.6.32 supports multiple VIFs.

Furthermore, this patchset allows wl12xx to use a single, updated firmware.bin to function in both AP and STA modes.

I'm currently rebuilding my build VM after spotty power these past few days messed with my images
(shouldn't have enabled write-caching for VMware!)

EDIT: Wow, that's a lot of commits.
I've pulled in loads of updates for wl12xx - AP mode should now be complete. In-kernel.

Pull in the newest linux-firmware to see it in action.

Now, on to more news:

I'm not yet moving onto even newer wl12xx firmwares as the next commit (SHA1: c8bde243421d759844264cf11e4248e7862c2722) would be incompatible with the shipped, on-device firmware.

BTW, debugfs functionality has been stripped out of wl12xx, it was causing build problems - if you want it back, bug me enough :p

What do you guys think, moving forward?

Re: [Announce] kernel-plus for Harmattan
 

Did you have to patch the kernel in the mmc core area to get the later tw wireless drivers to compile?

The newer tw stuff I have in a compat package needs stuff in kernel 2.6.37 minimum.

Re: [Announce] kernel-plus for Harmattan
 

I /was/ doing that, but then utterly phailed as mmc core update and ripping out 3.3's wl12xx driver was about one bajillion times harder to implement than backporting stuff from 2.6.39, I think.

Some of that mmc core stuff needs CONFIG_PM_RUNTIME, which would be nice to have on the N9 (selective device suspend, yay!), but after hours of patching wl12xx and mac80211, I'm going to take a nice break to preserve my sanity.

Anyone with more coffee than I do, please try to feel like cleaning up my commits if they get pulled into harmattan/kernel-plus-harmattan :D

Re: [Announce] kernel-plus for Harmattan
 

I'm going to try my hand at backporting some patches to first make the 2.6-compat package work for the tw wireless driver. I'm starting work on it this week given the US thanksgiving holiday.

If I get it working, I'll see if we can get the changes merged into this kernel branch if you'd like.

Re: [Announce] kernel-plus for Harmattan
 

Hey guys, just finally had time to test my recent changes.

Pelicularly, this new kernel /doesn't/ boot.

@juiceme, could you please load it up on your test N9 to see what's going on? Corresponding sources are at hurrian/kernel-plus-harmattan.

If it bombs somewhere related to mac80211/wl12xx, I'll revert the commits for now, if it's somewhere else guess I'll have to dig around a bit more.

Re: [Announce] kernel-plus for Harmattan
 

I'm glad to do that, but unfortunately it has to wait over the weekend... I left my rig at my workpalce and will be able to get to it on monday morning :o

Re: [Announce] kernel-plus for Harmattan
 

systemd requires the directory /sys/fs/cgroup as a mountpoint. It doesn't exist in the kernel (2.6.35). It can't be created with mkdir either because sysfs doesn't allow that. To create the directory this patch from the 2.6.36 kernel must be backported.

http://git.kernel.org/?p=linux/kerne...9dae0071bd11f3

The following kernel config options must be selected.
General setup --->[*] Control Group support
Device Drivers --->
Generic Driver Options --->[*] Maintain a devtmpfs filesystem to mount at /dev[*] Automount devtmpfs at /dev, after the kernel mounted the rootfs (NEW)
File systems --->
< > Kernel automounter support
<*> Kernel automounter version 4 support (also supports v3)

Re: [Announce] kernel-plus for Harmattan
 

I can confirm the device just hangs there, propably it is the sys mount that it is after. (did not try to patch it yet)

This is the last things that get written to console:

Re: [Announce] kernel-plus for Harmattan
 

hmm is this normal ?

Re: [Announce] kernel-plus for Harmattan
 

EDIT: Posted. See Page 1.

Sources patched, building.
kernel-plus 2.6.32.61-plus v5 will be built from the old local source directory, with source published as a .patch (again, sadly)

Rest assured, I'll try to match the kernel source with the Git source next time.

Details for the next release:

• systemd support [On Harmattan!]
  (cgroupfs: create /sys/fs/cgroup to mount cgroupfs on)
• /proc/atags exported
  Anyone want to work on porting uboot to N9/50? Start with this!
• CONFIG_KEXEC enabled as a requirement of atags export
• sched: Tweak sched_latency and min_granularity
• mutex: Improve the scalability of optimistic spinning
• net: __alloc_skb() speedup
• KSM is also enabled, but not by default. eglibc may need to be recompiled for MADV_*MERGEABLE support. Programs linking into eglibc may need to be rebuilt. Uh-oh.

Nokia internal dev stuff can boot off MMC, NAND or USB, IIRC.

You can see some remnants of this code on the N900.

Re: [Announce] kernel-plus for Harmattan
 

Looking forward, and very happy with the progress speed on this kernel :)
One thing.
I think we should include kernel drivers for Prolific PL2303 USB-to-Serial port adapter in the kernel
I believe this could possibly enable the option for USB serial console. Not quite as good as Juice's 'real' serial, but, nevertheless, usefull.

http://www.tldp.org/HOWTO/Remote-Ser...re-kernel.html

http://cateee.net/lkddb/web-lkddb/US...AL_PL2303.html

Re: [Announce] kernel-plus for Harmattan
 

Yes, when you have R&D mode active that is part of preinit that's executed.

Re: [Announce] kernel-plus for Harmattan
 

That would be useful, more so if you can initialize the USBSERIAL device as early as possible in order to capture all possible glitches.

Re: [Announce] kernel-plus for Harmattan
 

Hi

I am total noobie when it comes to this kernel stuff.

I have been following this thread from time to time and concluded that this probably isn't important for me as a basic user. I don't think that I would bee needing ExFAT or NTFS on my N9, I do get Injection from N900 days, but don't know if I need it either.

What I would be interested is that does this bring any performance/stability benefits? Besides Optimizer.

Really great to see that people are still working and tweaking the N9 under the hood. Thank you guys

Re: [Announce] kernel-plus for Harmattan
 

Well, of the top-of -my-head, stuff that are there or might soon be;
- iptables with all targets including full NAT
- wlan AP mode
- BT low energy mode
- possible performance enchancements
- bug corrections (FUSE performance, anyone?)
- all fancy new devices
- USB host mode
- etc. etc.

Re: [Announce] kernel-plus for Harmattan
 

btw, are you looked for latest bluez and modules backport?
https://thp.io/2013/maemo/n9_bh221.html

Re: [Announce] kernel-plus for Harmattan
 

Not sure how the bluetooth is configured in this kernel. But, here are some info on what we may want applied

BLuez 4.97 . Yeah, I have compiled it - lets go test

https://www.dropbox.com/s/r76h0tm3p1...4.97_armel.deb

Re: [Announce] kernel-plus for Harmattan
 

Re: [Announce] kernel-plus for Harmattan
 

I believe that this is the driver for a USB to serial stick, in which case won't work until the Hostmode-on-N9 thing gets done.

An alternative is this totally 1337 method of getting to the kernel buffer.
Perhaps support for copying the buffer could be added in the next ubiboot - it's at a fixed location, it just needs to be copied into a file.

Haven't checked yet. I was planning to update BlueZ all the way to the latest version to allow BT4.0LE, but I don't know if we still want the Nokia patches.

Re: [Announce] kernel-plus for Harmattan
 

It is doable, yes. The problem however is that after a kernel panic you need a hard reset to get to boot again, and this may corrupt the memory of the device. I am not sure how the poweroff on N9 works, though, it may even retain voltages to the memory chip?
If the voltage is cut it depends on the memory device on how long the old state is held in the device.

I remember doing some experiments with Commodore 64 when that device was state of the art; I had a picture in memory, and when I power-cycled it and redrew the memorymap after the device came back to life it was pretty much still there.
To completely get rid of the drawing, so that it was not recognizable any longer took a power cut of maybe 30 seconds!

On those days memory cells were however at least 3 orders of magnitude larger, which may affect your mileage :D

Can add opptimizer module startup at booting?

Re: [Announce] kernel-plus for Harmattan
 

Please dont :D

install attachment to load opptimizer module on demand by opptimizer ui. just cutted modules from package and patched loader to work with kernel-compiled modules

Re: [Announce] kernel-plus for Harmattan
 

Oh, those were the days......
Thats how I 'hacked' a copy of all my protected games ;)

Re: [Announce] kernel-plus for Harmattan
 

Getting myself a second N9 :) 64G, black, all-inclusive :) 100€, good price ?

Re: [Announce] kernel-plus for Harmattan
 

sweet price... half year ago i got 16GB for $200...

Re: [Announce] kernel-plus for Harmattan
 

more than a year ago 16GB for 100$ and the new one cost now ~200$

Re: [Announce] kernel-plus for Harmattan
 

:) of course, I will disassemble it :P want that serial console

Re: [Announce] kernel-plus for Harmattan
 

better modify sim tray with some connectors

Re: [Announce] kernel-plus for Harmattan
 

I was thinking something like that. Might 'use' the headphone jack hehe, if possible

Re: [Announce] kernel-plus for Harmattan
 

I suggest the hacked simtray thing. Used that extensively before acquired another device which is now using permanent wired&soldered conections.

It is fairly easy to make an adapter that slots into the simcard hole. Only one thing you need to be careful; make the adapter so that it has a slot on the middle, or make it so thin that it does not bend the actual sim-connectors in the device. Doing that will give you problems later if the sim contacts do not connect well to the card.

Re: [Announce] kernel-plus for Harmattan
 

Juice, if it's not too much to ask, what are the dimensions of the debug pads area? Saw the N9 with the physically hacked SIM tray; knowing how deep they go and how wide each pad is would help making an adapter :p

LPDDR SDRAM holds the data for quite some time - VRAM glitches on Nemo N900 once spewed out some remaining bitmaps from Maemo 5 ;)

What I'm concerned about though, is that ubiboot logging to the RAM buffer would overwrite whatever's remaining in RAM.

The copy code would have to be placed really early in boot, maybe to a procfs entry, subsequently copied when ubiboot-menus is launched.

Re: [Announce] kernel-plus for Harmattan
 

I'll have to measure those, the pads are fairly large ans I made my first adapter by just guessing at the dimensions and filing the bits until they fit nicely :)


That takes some studying but it should not be too difficult to ensure there's no overwriting. I had to shuffle the kernel bits around in the latest ubiboot kernel anyway to get the modules load under the 16M limit from the main kernel block.

---- edit -----

Something I thought of today triggered when I thought about the status led flickering in R&D mode... I think it just shows up CPU activity, right?

Well. why not make the status led function as serial output!
That'd be lot easier than using the AUX output, since you'd not need any fancy modulation schemes and the bandwidth would be eornomous :D

You'd just need a photodiode hookup that you'd place on top of the led, converting it directly to serial signal wired to PC :D:D

Re: [Announce] kernel-plus for Harmattan
 

What an awesome idea :) Shouldnt be to hard to implement. Well, kinda'

Re: [Announce] kernel-plus for Harmattan
 

@hurrian One thing. The iptables redirect () seems to have gone haywire in cyberspace again :)

http://talk.maemo.org/showpost.php?p...5&postcount=42

At least, again this message shows up e.g. in ettercap:

iptables: No chain/target/match by that name.

Which is connected to CONFIG_IP_NF_TARGET_REDIRECT

Re: [Announce] kernel-plus for Harmattan
 

Enabled [M]. See first post for details!

kernel-plus 2.6.32.61 r6 focuses on btrfs. It has been enabled as [Y] in the kernel.

Note that the disk format as used in the kernel probably matches 2.6.34 - ubiboot definitely has not been updated to support reading from this version of the btrfs disk format. Play with /home or something.
Meanwhile, or if you're feeling adventurous, load zImage from somewhere other than the Harmattan rootfs (check ubiboot.conf)