Screen freezes automatically after a while in Kobo Glo

[chandravadans]

I'm running the koreader version koreader-nightly-20140618 (latest nightly build as of June 23, 2014) on my Kobo Glo with firmware version 3.3.0

koreader automatically freezes after a few seconds of being idle on opening a book, and I have to restart to get everything back. Here's the content of crash.log

NtxHwCfg_GetCfgFldStrVal:[WARNING]Config version too old !! Please update config file !!

I've tried this on multiple pdfs, and the result is always the same. Would appreciate any help on this, as koreader is probably the best reader I've tried for pdf's on the kobo and I intend to use it for all my pdf needs :)

Steps to reproduce the problem :

1. Open a side loaded pdf in koreader
2. Leave it idle for a while without turning pages (for about 10s)

Result : Screen freezes and only thing that can be done is a reset.

Here is the detailed log. I opened a pdf, turned a page quickly, enabled reflow and waited for a while before attempting to turn the page again.

NtxHwCfg_GetCfgFldStrVal:[WARNING]Config version too old !! Please update config file !!
# key ev {
    ["type"] = 1,
    ["time"] = {
        ["usec"] = 471904,
        ["sec"] = 1403506350
    },
    ["code"] = 330,
    ["value"] = 1
}
# in tap state...
# set up hold timer
# in tap state...
# in tap state...
# in tap state...
# in tap state...
# in tap state...
# key ev {
    ["type"] = 1,
    ["time"] = {
        ["usec"] = 538985,
        ["sec"] = 1403506350
    },
    ["code"] = 330,
    ["value"] = 0
}
# in tap state...
# set up tap timer
# in tap timer true
# single tap detected in slot 0
# key ev {
    ["type"] = 1,
    ["time"] = {
        ["usec"] = 276004,
        ["sec"] = 1403506352
    },
    ["code"] = 330,
    ["value"] = 1
}
# in tap state...
# set up hold timer
# in tap state...
# in tap state...
# in tap state...
# key ev {
    ["type"] = 1,
    ["time"] = {
        ["usec"] = 334886,
        ["sec"] = 1403506352
    },
    ["code"] = 330,
    ["value"] = 0
}
# in tap state...
# set up tap timer
# in tap timer true
# single tap detected in slot 0
# key ev {
    ["type"] = 1,
    ["time"] = {
        ["usec"] = 823701,
        ["sec"] = 1403506354
    },
    ["code"] = 330,
    ["value"] = 1
}
# in tap state...
# set up hold timer
# in tap state...
# in tap state...
# in tap state...
# in tap state...
# in tap state...
# in tap state...
# in tap state...
# in tap state...
# key ev {
    ["type"] = 1,
    ["time"] = {
        ["usec"] = 916066,
        ["sec"] = 1403506354
    },
    ["code"] = 330,
    ["value"] = 0
}
# in tap state...
# set up tap timer
# in tap timer true
# single tap detected in slot 0
# key ev {
    ["type"] = 1,
    ["time"] = {
        ["usec"] = 574094,
        ["sec"] = 1403506356
    },
    ["code"] = 330,
    ["value"] = 1
}
# in tap state...
# set up hold timer
# in tap state...
# in tap state...
# in tap state...
# in tap state...
# key ev {
    ["type"] = 1,
    ["time"] = {
        ["usec"] = 670861,
        ["sec"] = 1403506356
    },
    ["code"] = 330,
    ["value"] = 0
}
# in tap state...
# set up tap timer
# in tap timer true
# single tap detected in slot 0
# key ev {
    ["type"] = 1,
    ["time"] = {
        ["usec"] = 743217,
        ["sec"] = 1403506358
    },
    ["code"] = 330,
    ["value"] = 1
}
# in tap state...
# set up hold timer
# in tap state...
# in tap state...
# in tap state...
# key ev {
    ["type"] = 1,
    ["time"] = {
        ["usec"] = 801884,
        ["sec"] = 1403506358
    },
    ["code"] = 330,
    ["value"] = 0
}
# in tap state...
# set up tap timer
# in tap timer true
# single tap detected in slot 0
# key ev {
    ["type"] = 1,
    ["time"] = {
        ["usec"] = 721666,
        ["sec"] = 1403506361
    },
    ["code"] = 330,
    ["value"] = 1
}
# in tap state...
# set up hold timer
# in tap state...
# in tap state...
# in tap state...
# in tap state...
# key ev {
    ["type"] = 1,
    ["time"] = {
        ["usec"] = 822522,
        ["sec"] = 1403506361
    },
    ["code"] = 330,
    ["value"] = 0
}
# in tap state...
# set up tap timer
# in tap timer true
# single tap detected in slot 0
# opening file /mnt/onboard/.pdf/Programming.Challenges-2003.pdf
# Loaded plugin plugins/evernote.koplugin
# init scroll page states
# going to page number 1
# init scroll page states
# setting zoom mode to page
# init scroll page states
# painting {
    ["y"] = 0,
    ["x"] = 0,
    ["h"] = 1003.8988095238,
    ["w"] = 758
} to 0 0
# rendering
# now painting {
    ["pageno"] = 1,
    ["excerpt"] = {
        ["y"] = 0,
        ["x"] = 0,
        ["h"] = 1003.8988095238,
        ["w"] = 758
    },
    ["size"] = 380541.64880952,
    ["bb"] = 
} {
    ["y"] = 0,
    ["x"] = 0,
    ["h"] = 1003.8988095238,
    ["w"] = 758
}
# hinting page 2
# painting {
    ["y"] = 0,
    ["x"] = 0,
    ["h"] = 1003.8988095238,
    ["w"] = 758
} to 0 0
# now painting {
    ["pageno"] = 1,
    ["excerpt"] = {
        ["y"] = 0,
        ["x"] = 0,
        ["h"] = 1003.8988095238,
        ["w"] = 758
    },
    ["size"] = 380541.64880952,
    ["bb"] = 
} {
    ["y"] = 0,
    ["x"] = 0,
    ["h"] = 1003.8988095238,
    ["w"] = 758
}
# key ev {
    ["type"] = 1,
    ["time"] = {
        ["usec"] = 755882,
        ["sec"] = 1403506365
    },
    ["code"] = 330,
    ["value"] = 1
}
# in tap state...
# set up hold timer
# in tap state...
# key ev {
    ["type"] = 1,
    ["time"] = {
        ["usec"] = 772400,
        ["sec"] = 1403506365
    },
    ["code"] = 330,
    ["value"] = 0
}
# in tap state...
# set up tap timer
# in tap timer true
# single tap detected in slot 0
# goto relative page: 1
# going to page number 2
# init scroll page states
# set page position 0

[WS64]

I had the same issue.
In my case it was on a Kobo Mini where I had removed the internal SD card and replaced it with a bigger one.
Although everything worked, Koreader (and actually also Start Menu) crashed after a minute or two (did not need more time) being idle and the reader always became inresponsible. No issues with that card in nickel.
I then replaced this card with another one (a 1:1 copy of the first card played back to a new one), and all crashes were gone.
In case you also exchanged the internal SD card, try another one, maybe this is the same issue as mine was.

[chandravadans]

Actually, I haven't gotten that far yet. Its the same internal SD card that came with the device. Thanks for letting me know, anyway :)

[hwhw]

Hm, nothing in the log indicates an error. It looks a bit as if Koreader simply doesn't get any more input events. I'm not sure how to debug this, unfortunately.

[chandravadans]

Found a fix. More of a work around, actually. Have no clue as to why it works, can only speculate :P

• Enable usbnet and do the following (Basically, ssh into your kobo)

$ sudo ifconfig usbnet0 192.168.2.100
$ ssh root@192.168.2.101

when prompted for a root password, just hit return. There's no password.

• Then, cd into mnt/onboard/.kobo/koreader, and launch koreader.

[root@(none) ~]#  cd mnt/onboard/.kobo/koreader
[root@(none) ~]# ./koreader.sh 

• koreader starts on the device. That's it. No more 'waiting crashes'. Ever.

It was interesting to note the following error message on initiating the ssh connection

[1185] Jun 23 12:58:48 lastlog_perform_login: Couldn't stat /var/log/lastlog: No such file or directory
[1185] Jun 23 12:58:48 lastlog_openseek: /var/log/lastlog is not a file or directory!

Maybe that was the 'root' of the problem. But then again, maybe :)

I'll probably leave the issue 'open' till I figure out what was wrong, why the fix worked, and if the trouble was with koreader or kobo :)

[hwhw]

It should not be related - well, it's always a bit mysterious with proprietary software running alongside :-)

Does it work even if there is no USB connection anymore? Because a USB connection will keep some things up and running and the device might behave differently.

Maybe some other software is triggered after that time and pulls the input events out from under us.

[chandravadans]

When I disabled 'usbnet', it started crashing again. I did not have to connect my device to the laptop or ssh into it, though. Just having it enabled did the trick . It crashes unless launched via ssh once per session.
I guess the solution is to look at what usbnet actually does to prevent the crashes from happening.

PS. I'm talking about the usbnet that comes with 'Sergeys Launcher'. Installing this is detailed here :

http://www.mobileread.com/forums/showthread.php?t=226456

It installs usbnet along with Sudoku, Calculator and a host of other distractions.

• Also, I don't seem to be able to mount the device for storage when this mode is enabled :/

[Markismus]

Yesterday we changed the koreader.sh script for kobo's. The commit #662 will alleviate a lot of problems with Nickel for firmware 3.3.0 and above. It was intended to fix the problem that all EV_KEY events, such as light button, sleep cover and power switch, were accumulated and only processed when Nickel continued.

So just copy-paste the linked content to your script or wait for the next nightly release. I am quite interested to hear whether your problem persists! It would mean that the event accumulation can be more widespread than only the EV_KEY events.

[WS64]

Again this fits to my problem, when USB connection was there no crashes. Maybe the internal card in your Kobo is like the one I replaced first, too slow/fast/hungry/whatever and should be replaced.

[chandravadans]

@Markismus : I did paste the modifications to koreader.sh and that didn't seem to make a difference. The problem still persists. There's nothing in the debug logs either. Just the line about the config being too old.

I haven't tried the option suggested by @WS64. Could you please detail how you replaced the cards? You just copied the folders into the new card? Or used dd? If its through dd, it shouldn't matter much, imho.
Also, I'm interested in knowing if even you use Calibre to manage your ebook collection? Maybe that's the problem?

[Markismus]

Thanks for trying!
So it isn't a problem with Nickel.
Does coolreader of Sergey run into the same problems? If not than it's a koreader problem.

Do the problems also arise when you open and read an epub? Do they arise when you read a dvju-file? All three file-types use a different backend to read the files.

It sounds like your device goes to sleep. Can you wake it up now you have th e new script with a pull of the power switch? Do you have a sleep cover? Does it also stop functioning without the sleep cover.

Could you see what the suspend commands do on your device? What response does echo 1 > /sys/power/state-extended and echo mem > /sys/power/state generate?

[chandravadans]

• No, Coolreader doesn't have this problem.
• Yes, the problem occurs with epubs too. It freezes after a while (though it was a wee bit longer before it froze with an epub). I haven't tried djvu documents, but I guess it would appear there too. I'll check it up and let you know.
• Nope, pulling the power switch doesn't bring it back. The only thing that does is a pin reset.
• No, I don't own a sleep cover. All these problems are without the sleep cover.

echo 1>/sys/power/state-extended did nothing, and
echo mem > /sys/power/state returned sh: write error: Device or resource busy
cat /sys/power/state gave standby mem, though. The device was not asleep. Is that as it you'd expect?

[WS64]

For me again no problem with Coolreader, just Koreader and Start Menu itself, did you try this one too?

And no, not just copy the partition. You need to remove the card, use a cardreader and copy all partitions (2 linux, one FAT32), write the whole thing 1:1 back to another card. optional you then can increase the size of the FAT32 partition in case you used a bigger card). No idea how to do it in linux, in windows you can use e.g. WinImage for the copying and EaseUs Partition Master to change the size of the FAT32 partition.
When all is done put the new card into the Kobo and you are done.
Here's a bit more about it: http://www.mobileread.mobi/forums/showthread.php?t=209122

I use Calibre for my collection, but not anymore to fill books to my Kobo.

[chandravadans]

No, I haven't installed start menu. Personally, having to reboot just to switch the reader seems a little uncomfortable to me. Anyway, thanks for the description, @WS64 ! I shall give this a try when possible and get back :)

Meanwhile, still stumped about the 'sleep crash' issue. It seems to occur in almost all formats supported by koreader.

[WS64]

With Start Menu you can very easy switch between Nickel and Koreader or Coolreader, I doubt you are faster in any other way!

And Koreader, mine used to crash even when you stay in the menu, no book open, and wait for a minute...

[Markismus]

There is indeed no need to reboot in Start Menu between Nickel and koreader. However, when you exit koreader for the first time to enter Nickel, you will see dancing squares. This is not a reboot, but the first start of Nickel after a shutdown. The second switch will be almost instantaneous, especially with the new koreader-Nickel loop scripts Tshering created for koreader users!

Since the new script didn't affect your problem, I don't expect much from Start Menu, since it is the same code. Still Start Menu is very nice for hacking...

[chandravadans]

@Markismus : Yes.. Even startmenu's crashing. How about Sergey's launcher code, then? It seems to be pretty stable, and doesn't sleep crash. Maybe we could have a look at what it does to keep it from not doing so?

@WS64 : Yes, sleep crashing in the menu confirmed. :( I've ordered a new microSD card, and am waiting for it to arrive, so I can test your solution :) Btw, was yours a class 10 microSD?

[WS64]

The one I have in now and another one that is also working are both class 4. I read somewhere that someone said his Kobo does not work with class 10, but of course that might just be coincidence. I can check later at home if the bad one was a class 10, I actually think so. It was for sure brand Hama, the working ones are all SanDisk.

[chandravadans]

Update.
I installed an earlier version of koreader (koreader-kobo-arm-linux-gnueabihf-v2014.04-65-g24bc3) , and now the contents of crash.log are:

NtxHwCfg_GetCfgFldStrVal:[WARNING]Config version too old !! Please update config file !!
error: invalid page object
error: invalid page object

also, theres a stack_00.log in the .kobo directory, which says

OH THE HUMANITY!
pid: 1285, tid: 1285, rev: 8eb9c1f8b250c6e47ce190d41a8b7d67de34f31e
  >>> /usr/local/Kobo/nickel <<<
signal 6 (SIGABRT), code -6 (?), fault addr --------
 r0 00000000  r1 00000505  r2 00000006  r3 00000505
 r4 00000006  r5 2f1ec5c0  r6 2f1ad08c  r7 0000010c
 r8 7e9f3870  r9 7e9f3a34  10 00000008  fp 005cfef8
 ip 0000010c  sp 7e9f3694  lr 2f0f2e83  pc 2f0e5276  cpsr 20000030
    #00 sp: 0x7e9f3694 ip: 0x2f0e5276  /lib/libc-2.11.1.so: gnu_get_libc_version+0x195
    #01 sp: 0x7e9f369c ip: 0x2f0f2e83  /lib/libc-2.11.1.so: gsignal+0x2a
    #02 sp: 0x7e9f36a4 ip: 0x00000005  : +0x2a
unw_get_proc_info(ip=0x5) failed: ret=-10

looks like a pretty bad segfault. I wonder what it means, and if it helps in zeroing on the root of the problem.

[Markismus]

"Oh the humanity errors" are from Kobo (and Nickel).

The non-working Start menu...did you perchance install Giorgio's multitouch hack? It is known no to work with Tshering's Start menu.

[chandravadans]

No, I haven't installed the multi touch hack. Do the invalid page object errors mean anything? They seem to be occurring consistently across all versions of koreader!

[Markismus]

It's a Nickel crash...those happen...independently from koreader...I had some, but never bothered with them, because it's Nickel...Googling libc-2.11.1.so..c-library for system calls...apparently Ubuntu users suffer from it, too. Someone there fixed it with rm -rf /var/cache/apt/*.bin...

Since Start Menu fails to work too, you could ask Tshering for some support.

[chandravadans]

Yes, I have. For future reference, the post is at http://www.mobileread.com/forums/showpost.php?p=2858195&postcount=259
Thanks for your support! :)

[chandravadans]

Got a fix for the startmenu 'sleep crash' problem. Apparently the crashes don't occur if the display is kept busy (by displaying the clock, in startmenu's case). Maybe something of the sort could be done for koreader as well?

@WS64 : I tried a SanDisk 16GB Class 10 card. The crashes continue to happen :(
I'll try to get my hands on a Class 4 SD card if possible.

[chandravadans]

FIXED the sleep crash issue. It works well without the changes in koboreader.sh made in commit #664. With the changes, the device just restarts on exit. The changes are here. Made a pull request.

[Markismus]

Could you log the output of the koreader script? I am quite surprised that the device restarts.

You could change the fmon koreader call to:
"koreader.sh 2>1 >koreader.sh.log" ?

[chandravadans]

I've just tried out some tests and here are the results on Kobo Glo, firmware version 3.3.0:

• Newer koboreader.sh, without adding my script, exits cleanly. It also sleep crashes.
• Newer koboreader.sh, after adding my script doesn't sleep crash, but restarts when I try to exit from koreader. I get a OH THE HUMANITY error and an ugly segmentation fault. Modifying the fmon koreader launcher didn't produce the file koreader.sh.log. I guess because it crashed.
• Older koboreader.sh, without adding my script sleep crashes.
• Older koboreader.sh, after adding my script, doesn't sleep crash and exits to nickel cleanly.
• I can't suspend or poweroff the device using the power button with either of the versions. The events get buffered and the device suspends after exiting koreader. So I exit koreader completely and then pull the switch to suspend/poweroff. I haven't checked if the device auto suspends automatically after the default timeout, though. My guess is it wouldn't as nickel is killed off in both the versions.

[houqp]

hm... just adding your wile true write file script should not cause a segfult.

In fact, this "keep busy script" can be simulated in koreader. Just for testing purpose, can you try adding following code at the beginning of reader.lua (around line 39)?

function loop(s)
    local f_out = io.open("/tmp/test", "w")
    f_out:write("1")
    f_out:close()
    UIManager:scheduleIn(1, function() s(s) end)
end
loop(loop)

[Markismus]

I was able to reproduce the crash. Terminal output:

killall: nickel: no process killed
killall: hindenburg: no process killed
NtxHwCfg_GetCfgFldStrVal:[WARNING]Config version too old !! Please update config                      file !!
NtxHwCfg_GetCfgFldStrVal:[WARNING]Config version too old !! Please update config                      file !!
Executing fmon script: /mnt/onboard/.kobo/fmon/launcher.sh
Executing fmon script: /mnt/onboard/.kobo/fmon/koreader.sh
Error: invalid arguments!
This application failed to start because it could not find or load the Qt platform plugin "xcb".

Available platform plugins are: kobo.

Reinstalling the application may fix this problem.
"/usr/local/Kobo/nickel() [0x181f6]"
"/usr/local/Kobo/nickel() [0x185c2]"
"/usr/local/Kobo/nickel() [0x12d0a]"
"/lib/libc.so.6(__default_sa_restorer_v2+0) [0x2f0f3aa0]"
"/lib/libc.so.6(+0x17276) [0x2f0e5276]"
"/lib/libc.so.6(gsignal+0x29) [0x2f0f2e82]"

[chandravadans]

killall: nickel: no process killed
killall: hindenburg: no process killed

That's interesting!
This was done with the new koboreader.sh, and with my script?

[Markismus]

The reason everything worked was that I still used the old script koreader_kobo.sh on my mom's Glo. Can't see that from the screen.

Those killall messages are due to the fact that i ran koreader.sh multiple times, without exiting koreader properly. Nickel is then never restarted or continued.

[Markismus]

Clean repeat:

[root@(none) koreader]# ./koreader.sh
NtxHwCfg_GetCfgFldStrVal:[WARNING]Config version too old !! Please update config file !!
NtxHwCfg_GetCfgFldStrVal:[WARNING]Config version too old !! Please update config file !!
Executing fmon script: /mnt/onboard/.kobo/fmon/launcher.sh
Executing fmon script: /mnt/onboard/.kobo/fmon/koreader.sh
fmon started!
This application failed to start because it could not find or load the Qt platform plugin "xcb".

Available platform plugins are: kobo.

Reinstalling the application may fix this problem.
"/usr/local/Kobo/nickel() [0x181f6]"
"/usr/local/Kobo/nickel() [0x185c2]"
"/usr/local/Kobo/nickel() [0x12d0a]"
"/lib/libc.so.6(__default_sa_restorer_v2+0) [0x2f0f3aa0]"
"/lib/libc.so.6(+0x17276) [0x2f0e5276]"
"/lib/libc.so.6(gsignal+0x29) [0x2f0f2e82]"

[chandravadans]

@houqp : The device restarted. I put the code right before function exitReader(). Was it where you wanted it to be put?
It segfaulted with the OH THE HUMANITY error.

[chandravadans]

@Markismus : that's the new version of koreader.sh, I take it? With or without the script in the background?

[houqp]

@Markismus is this the log for sleep crash?

@chandravadans yes. did the device restart when you run koreader or after you exit from koreader?

[chandravadans]

@houqp It happened as soon as I clicked on the koreader icon on the device. I haven't tested out the ssh way of launching it though. Should I?

[hwhw]

I don't think it has anything to do with the snippet about writing to a file. It's just the difference between killing nickel vs stopping nickel. The restart of nickel after killing it, to be most specific. It is wrong. Isn't there a proper way to stop and start nickel? The setting of the QWS_* environment variables also looks veeeeeery fishy to me. Would this be the same on all Kobo platforms? Then, however, I can't see how this would cause the freezes. I only see how this would cause a restart.

[Markismus]

Correct. The freezes happen with both the older koreader_kobo.sh and the new koreader.sh. The restart error emerges only with the newer script.

The fishy code was taken from Tshering's Start Menu. He has derived most of his code from adapting the \etc\init.d\rcS file from Kobo. For example line 22-31 in koreader.sh is taken from line 128-138 from rcS omitting the line echo 1 > /sys/devices/platform/mxc_dvfs_core.0/enable.

[houqp]

@chandravadans yes, test out in ssh should give us a crash log hopefully.

@hwhw , but both ways (kill and stop) freezes the screen without that while loop snippet. I suspect the device goes into freeze state when the system is idle? Then it's strange that there no other process writing files in the backgroud, for instance a logger.

[houqp]

@Markismus the new script does not restart without @chandravadans 's keep busy script right?

[chandravadans]

It does restart. With the old one, it doesn't.

[Markismus]

@houqp With your code snipped build 229 on Kobo Glo Firmware 3.3.0 does not freeze. On 3.1 it didn't freeze at all.
We should open another issue for the restart error. I'll first install Start Menu to see whether I can reproduce Chandravadans' error there and discuss it with Tshering.

Simply reverting back to the old script would mean that the Glo cannot use it's power button and light button. The script works on the Aura. Like most things with Kobo, it sadly seems device depend.

[houqp]

so yeah, these are two separate issues. for the freeze issue, we need to find out the maximum idle time and use that to replace 1s.

also this loop should be disabled on suspend to save power.

[chandravadans]

@Markismus : I tried it out on v4 of the start menu, and the sleep crash problem did exist. I reported it on the fourms and tshering had replied here It was his comment that gave me the idea of putting the busy wait. :)

• Also, I'm not able to use the power and the light buttons on the Glo even with the newer version of koreader.sh. Could you test it out on your Glo and let me know?

[Markismus]

It seems the problem is not new as #576 already discusses it.

After about 20 minutes the Glo went into standby and rebooted. The koreader.log is clear. Where have all the normal debug messages gone? Only new thing is this reoccurence:

# key ev {
    ["type"] = 1,
    ["time"] = {
        ["usec"] = 912541,
        ["sec"] = 1403691868
    },
    ["code"] = 330,
    ["value"] = 0
}
#

Is this the loop?

After 2 hours it quit and rebooted again. So there's a bigger problem with Glo on 3.3.0 than is solved with the loop

[houqp]

Do you get that key event consistently? I assume you get this with the loop hack? It is not the loop.

[Markismus]

In the small log that I generated it occured at regular intervals throughout the log.

[chandravadans]

I don't think it has anything to do with the loop.. Even i got it before.. Its attached in my main post, if you observe..

[pazos]

@chandravadans Could you check if you're are able to back to nickel after applying changes on #689 ??

I'm in firmware 3.4.1 on a kobo mini and experiment no hangs in both koreader and nickel (I'm using fmon to trigger koreader.sh) Suspend works as expected

[chrox]

Seems solved. Close this now.

[chandravadans]

Yes, I just tested it and it is confirmed that the present release of koreader solves the problem for kobo glo, running firmware 3.3.0. Suspend and light buttons work, and it doesn't freeze when idle.
Thanks for the support! :)

Ps. Sorry for the delay, been a little busy off late.