Active Topics
-
Firefox with Leste (5)
to Maemo 7 / Leste by teroyk - 1 day, 2 hrs ago -
Handbook (2)
to Maemo 7 / Leste by teroyk - 2 days, 1 hr ago - more...
| The Following 8 Users Say Thank You to freemangordon For This Useful Post: | ||
|
2012-04-20
, 19:00
|
|
Posts: 237 |
Thanked: 274 times |
Joined on Jul 2010
|
#142
|
Originally Posted by freemangordon
I lack experience for properly tracking the cause, if you provide some pointers on what logs should be collected I can happily provide them.
Well, we have NFC why those are crashing, if it is a SIGILL crash, then thumb errata workaround is not working and we will have to pray Nokia for a new Secure Monitor, which sets ACTLR.IBE bit.
On the other hand it is better to gather some logs first, analyze them and make conclusions.
| The Following 4 Users Say Thank You to ivyking For This Useful Post: | ||
|
|
2012-04-20
, 19:16
|
|
Posts: 2,076 |
Thanked: 3,268 times |
Joined on Feb 2011
|
#143
|
Originally Posted by ivyking
Awesome, thanks a lot. Only one point left that will probably cause me headaches tonight: boot.scr, could you upload yours (assuming it is same on all devices after following your steps, if not short description of how it should be created would be very much welcome)
yes I fixed the "--no-log" part in the script then installed, selected 12.04 daily, old ubuntu kernel, local firmware (copied from maemo), all dialogs with default selections.
after finishing:
extract unknown's .deb, copy the uImage to the first partition of the sdcard, copy the 2.6.37-custom folder to /lib/modules/
create a new boot.scr to load the uImage.
finally do "depmod -a" for modules to load correctly.
that's all, please do test it yourself it doesn't crash that much really..
|
|
2012-04-20
, 19:28
|
|
Posts: 2,076 |
Thanked: 3,268 times |
Joined on Feb 2011
|
#144
|
Originally Posted by freemangordon
Just a wild guess: could those crashing apps rely on some 3.0(3.2?) kernel features while being ran on 2.6.37? From the report core seems stable and only some apps crashing, hope this is it. If I manage to get it running will provide logs if you specify which/where/how to obtain them. This is very promising.
Well, we have NFC why those are crashing, if it is a SIGILL crash, then thumb errata workaround is not working and we will have to pray Nokia for a new Secure Monitor, which sets ACTLR.IBE bit
On the other hand it is better to gather some logs first, analyze them and make conclusions.
EDIT:
If i have time tomorrow I will enable the errata in KP and will modify it to dump the contents of ACTLR, so I will be able to check if we have access to that bit from non-secure mode. If we can't change it, the thumb errata does nothig and we will have to pester and pray for a new Secure Monitor. If the bit can be changed, then arm thumb errata works as it should, but we have another problem yet to be investigated.
@ivyking - any chance to gather some logs and/or coredump?
| The Following 5 Users Say Thank You to szopin For This Useful Post: | ||
|
|
2012-04-20
, 19:31
|
|
Posts: 237 |
Thanked: 274 times |
Joined on Jul 2010
|
#145
|
Originally Posted by szopin
should be the same as nemo's, I just named the kernel uImage to unknown-uImage to differentiate it from other kernels.
Awesome, thanks a lot. Only one point left that will probably cause me headaches tonight: boot.scr, could you upload yours (assuming it is same on all devices after following your steps, if not short description of how it should be created would be very much welcome)
setenv atagaddr $init_atagaddr
setenv bootcmd 'mmc init; fatload mmc 0:1 0x82000000 unknown-uImage; bootm 0x82000000'
setenv bootargs 'root=/dev/mmcblk0p3 rootwait console=tty0 omapfb.vram=0:2M,1:2M,2:2M mtdoops.mtddev=2 nosplash elevator=noop'
boot
if you are going to use the attached boot.scr, rename the uImage in the .deb to unknown-uImage.
remove the .txt extension ofcourse
[Edit] please note that this has been tested with old u-boot, not pali's version.
Last edited by ivyking; 2012-04-20 at 19:38.
| The Following 7 Users Say Thank You to ivyking For This Useful Post: | ||
|
|
2012-04-21
, 03:03
|
|
Posts: 58 |
Thanked: 197 times |
Joined on Apr 2012
|
#146
|
I am not sure about that install script in the actual condition.
I mean, you'd have to run in on a ubuntu 12.04 machine or at least you have to make sure to use a qemu-user-static newer then the one provided in ubuntu 11.10 repositories. To support hard-float. Seems to be a wrong statement. I read a bug report somewhere but can't find anything regarding floatpoint in qemu changelog.
I am also not sure about the integrations package that installs. Like I stated before, something doesn't work right and should be checked first in perspective to this ubuntu and kernel combination.
Also elevator=noop in boot.scr.... I am also not sure about that.
Last edited by unknown.obvious; 2012-04-21 at 03:27.
I mean, you'd have to run in on a ubuntu 12.04 machine or at least you have to make sure to use a qemu-user-static newer then the one provided in ubuntu 11.10 repositories. To support hard-float. Seems to be a wrong statement. I read a bug report somewhere but can't find anything regarding floatpoint in qemu changelog.
I am also not sure about the integrations package that installs. Like I stated before, something doesn't work right and should be checked first in perspective to this ubuntu and kernel combination.
Also elevator=noop in boot.scr.... I am also not sure about that.
Last edited by unknown.obvious; 2012-04-21 at 03:27.
|
|
2012-04-21
, 05:59
|
|
Posts: 237 |
Thanked: 274 times |
Joined on Jul 2010
|
#147
|
Originally Posted by unknown.obvious
elevator=noop specifies the disk io scheduler, it can be left out (cfq will be used as default) and can be changed after the system has booted.
I am not sure about that install script in the actual condition.
I mean, you'd have to run in on a ubuntu 12.04 machine or at least you have to make sure to use a qemu-user-static newer then the one provided in ubuntu 11.10 repositories. To support hard-float. Seems to be a wrong statement. I read a bug report somewhere but can't find anything regarding floatpoint in qemu changelog.
I am also not sure about the integrations package that installs. Like I stated before, something doesn't work right and should be checked first in perspective to this ubuntu and kernel combination.
Also elevator=noop in boot.scr.... I am also not sure about that.
elevator=noop is used because i/o is actually done through memory card controller which does its own optimizations ?
| The Following 4 Users Say Thank You to ivyking For This Useful Post: | ||
|
|
2012-04-21
, 15:52
|
|
Posts: 3,074 |
Thanked: 12,960 times |
Joined on Mar 2010
@ Sofia,Bulgaria
|
#148
|
@unknown.obvious - could you please make the following change, rebuild an upload the resulting uImage:
drivers/mmc/host/omap_hsmmc.c change:
#define OMAP_MMC_MAX_CLOCK 52000000
to
#define OMAP_MMC_MAX_CLOCK 48000000
that seems to me as a possible reason for having -110 error with some sdcards, as in stock n900 kernel the max frequency is 48MHz
drivers/mmc/host/omap_hsmmc.c change:
#define OMAP_MMC_MAX_CLOCK 52000000
to
#define OMAP_MMC_MAX_CLOCK 48000000
that seems to me as a possible reason for having -110 error with some sdcards, as in stock n900 kernel the max frequency is 48MHz
| The Following 8 Users Say Thank You to freemangordon For This Useful Post: | ||
|
|
2012-04-21
, 17:49
|
|
Posts: 58 |
Thanked: 197 times |
Joined on Apr 2012
|
#149
|
Originally Posted by freemangordon
Sorry. There is no OMAP_MMC_MAX_CLOCK in omap_hsmmc.c in 2.6.37 meego kernel source.
@unknown.obvious - could you please make the following change, rebuild an upload the resulting uImage:
drivers/mmc/host/omap_hsmmc.c change:
#define OMAP_MMC_MAX_CLOCK 52000000
to
#define OMAP_MMC_MAX_CLOCK 48000000
that seems to me as a possible reason for having -110 error with some sdcards, as in stock n900 kernel the max frequency is 48MHz
|
|
2012-04-21
, 17:52
|
|
Posts: 3,074 |
Thanked: 12,960 times |
Joined on Mar 2010
@ Sofia,Bulgaria
|
#150
|
Originally Posted by unknown.obvious
Could you give me a hint where the kernel source is located, sorry for the dumb question
Sorry. There is no OMAP_MMC_MAX_CLOCK in omap_hsmmc.c in 2.6.37 meego kernel source.
On the other hand it is better to gather some logs first, analyze them and make conclusions.
EDIT:
If i have time tomorrow I will enable the errata in KP and will modify it to dump the contents of ACTLR, so I will be able to check if we have access to that bit from non-secure mode. If we can't change it, the thumb errata does nothig and we will have to pester and pray for a new Secure Monitor. If the bit can be changed, then arm thumb errata works as it should, but we have another problem yet to be investigated.
@ivyking - any chance to gather some logs and/or coredump?
Last edited by freemangordon; 2012-04-20 at 18:59.