bugmldonkey, a multi-networks file-sharing client - Bugs: bug #17041, Reassigned to another tracker...

 
 

You are not allowed to post comments on this tracker with your current authentication level.

bug #17041: Reassigned to another tracker [was: On arm: hangs with 0% CPU until I strace it]

Submitter:  Samuel Mimram <smimram>
Submitted:  Thu 06 Jul 2006 06:21:56 AM UTC
   
 
Category:  Core Severity:  4 - Important
Item Group:  Program malfunction Status:  Postponed
Assigned to:  None Open/Closed:  Closed
Release:  2.7.7 Release: 
Operating System:  Linux Binaries Origin:  CVS / Self compiled
CPU type:  ARM

Jump to the original submission

Sun 18 Feb 2007 12:17:25 AM UTC, comment #23: 

THIS ITEM WAS REASSIGNED TO PATCH #5752.

Please, do not post any new comments to this item.


spiralvoice <spiralvoice>
Group administrator
Sun 18 Feb 2007 12:17:15 AM UTC, comment #22: 

Updated patch to remove option changes from patch #4768

(file #12004)

spiralvoice <spiralvoice>
Group administrator
Wed 31 Jan 2007 11:36:25 AM UTC, comment #21: 

I found a collection of porting infos from aMule project
specifically for ARM platform. I am not a C/C++ crack
but it may be possible that MLDonkey suffers from
similar problems.

http://forum.amule.org/thread.php?threadid=4676
http://forum.amule.org/thread.php?threadid=7502
http://forum.amule.org/thread.php?threadid=7645
http://lists.debian.org/debian-gcc/2001/04/msg00102.html

spiralvoice <spiralvoice>
Group administrator
Fri 21 Jul 2006 03:02:42 PM UTC, comment #20: 

@kevin: I thought so that arm277.patch (0. -> 0.01) does not solve
the bug because. Samuel now released 2.7.7-5 with the patch
removed which broke 2.7.3 for you so this core should now work.
On ARM the loop_delay function in 2.7.7-5 is the same, working one
as in 2.7.2, therefore I reduced this bug from blocker to important.
On all other platforms this bug seems to to exist.

Please report if the new Debian package does work for you.

spiralvoice <spiralvoice>
Group administrator
Fri 21 Jul 2006 03:00:07 PM UTC, comment #19: 

Hi!

I am afraid the patch on 2.7.7 does not fix the bug. The core is subject to the
 exact same freeze again.

I have not tested 2.7.7-5 yet but I am now compiling it. I suspect that it is
buggy. If 2.7.7-5 behaves the way I suspect, I will reopen debian bug #370107.

cheers
  Kevin

spiralvoice <spiralvoice>
Group administrator
Fri 21 Jul 2006 02:59:55 PM UTC, comment #18: 

Changes for Debian package:

Changes:
 mldonkey (2.7.7-5) unstable; urgency=low
   * Added arm_dynamic_loop_delay.dpatch on arm only to avoid a freeze of
     mldonkey, closes: #370107

spiralvoice <spiralvoice>
Group administrator
Thu 20 Jul 2006 03:39:16 PM UTC, comment #17: 

@Samuel: Ok, pango and I, we can currently not find the source of
this bug, which occurs only on ARM platform.

Please include arm.patch as ARM-specific patch in your Debian
package, after successful testing you can then close the bug
on Debian bug tracker, I will keep this bug open for some time
ahead, maybe we can come up with a fix later.

spiralvoice <spiralvoice>
Group administrator
Thu 20 Jul 2006 12:04:06 PM UTC, comment #16: 

Hi!

The 2.7.7 with the patch hangs again. loop_delay is at 20. It was at 20 when I tested the patched 2.7.3.

Just to make sure, I'm now again compiling
<http://prdownloads.sourceforge.net/mldonkey/mldonkey-2.7.7.tar.bz2?download>
with <http://savannah.nongnu.org/bugs/download.php?file_id=10371>

If this fixes it, I must have failed to patch 2.7.7 properly. If it hangs again,
Than either the patch doesn't do the same change to mldonkey as the 2.7.3 patch
OR the patch cured only one cause of trouble and the actual problem is more complex.

cheers
  Kevin

spiralvoice <spiralvoice>
Group administrator
Wed 19 Jul 2006 11:17:48 AM UTC, comment #15: 

@Kevin: Nonetheless would it be nice if you test arm277.patch, this would be a generic patch for all platforms. Its working here on x86 for > 12h now without problems.

@Samuel: If you want a quick fix now please use arm.patch

spiralvoice <spiralvoice>
Group administrator
Wed 19 Jul 2006 10:38:19 AM UTC, comment #14: 

I can (and will) include this patch in Debian on arm only.

Samuel Mimram <smimram>
Tue 18 Jul 2006 06:21:59 PM UTC, comment #13: 

Thanks for testing, good that we located the buggy code section.
Attached to this bug report, thanks to pango, you will find
arm277.patch, please apply it to MLDonkey 2.7.7 and, sorry,
compile again. Please raise loop_delay to the old default of 20
again before testing this version.

I would like to fix bug for you without a specific patch for
Debian ARM, there is no guarantee that this bug does not affect
other systems as well. But if a general solution can not be found
my question to the package maintainer is if its possible for
Debian to include a platform-specific patch.

spiralvoice <spiralvoice>
Group administrator
Tue 18 Jul 2006 05:58:35 PM UTC, comment #12: 

Frmo OP:

Hi!

Bug found: The patch fixed 2.7.3 -- it is now running properly. Thanks for finding the bug. Well done.

Would you like me to help getting Dynamic loop delay to work on arm or do you
want to turn it off on that arch completely i.o.t. close the bug?

cheers
  Kevin

spiralvoice <spiralvoice>
Group administrator
Tue 18 Jul 2006 05:58:04 PM UTC, comment #11: 

From OP:

Hello!

Playing with delay_loop has not affected the freeze. The value was at 20
(initially) then 0, 10, 30, 50 and every time the core froze a few minutes after
starting.

I have now successfully applied the patch and the patched 2.7.3 is now
compiling. Please give my NSLU2 further patience. O will let you know.

cheers
  Kevin

spiralvoice <spiralvoice>
Group administrator
Sat 15 Jul 2006 05:13:14 PM UTC, comment #10: 

Mail from OP:

Hi!
The cpu is de-underclocked from 133 to 266 MHz, giving it 266 bogomips.
(lpj=1331200)

Decreasing loop_delay from 20 to 1 does not affect the bug. I'm now increasing
it and watching, next test is the dynamic_loop_delay patch.

cheers
  Kevin

snip --------------------
cat /proc/cpuinfo
Processor       : XScale-IXP42x Family rev 1 (v5l)
BogoMIPS        : 266.24
Features        : swp half thumb fastmult edsp
CPU implementer : 0x69
CPU architecture: 5TE
CPU variant     : 0x0
CPU part        : 0x41f
CPU revision    : 1
Cache type      : undefined 5
Cache clean     : undefined 5
Cache lockdown  : undefined 5
Cache format    : Harvard
I size          : 32768
I assoc         : 32
I line length   : 32
I sets          : 32
D size          : 32768
D assoc         : 32
D line length   : 32
D sets          : 32

Hardware        : Linksys NSLU2
Revision        : 0000
Serial          : 0000000000000000
snap --------------------

spiralvoice <spiralvoice>
Group administrator
Sat 15 Jul 2006 10:37:10 AM UTC, comment #9: 

How many bogomips has your ARM cpu? -> cat /proc/cpuinfo

spiralvoice <spiralvoice>
Group administrator
Sat 15 Jul 2006 10:35:43 AM UTC, comment #8: 

Before testing the attached patch please play around with
option loop_delay in downloads.ini, default should be
20 for you, try lowering it to 1, then increase it step
by step until, lets say, 50 and test if MLDonkey still freezes.

One possible source for the bug could be this patch:

2006/01/06
4768: Dynamic loop delay (pango)

Attached to this bug report is a patch which removes this
patch from MLDonkey 2.7.3 source code, you also have to
change downloads.ini option loop_delay back to 5 and hope
2.7.3 will function like older versions for you.

spiralvoice <spiralvoice>
Group administrator
Sat 15 Jul 2006 10:31:00 AM UTC, comment #7: 

Mail from OP:

Hello Spiralvoice!

Here my results:

1. The bug was definitely introduced with 2.7.3.
2. The logfile:

snip-----------------------------------
tail -n21 mlnet.log
   0    0    0    0    0    0    0    0    2    0    0    2     ready  2 active
   0    0    0    0    0    0    0    0    2    0    0    2     throttled ready
   0    0    0    0    0    0    0    0    1    0    0    1     indirect
None
KNOPPIX_V5.0.1CD-2006-06-01-EN.iso.md5
None
KNOPPIX_V5.0.1CD-2006-06-01-EN.iso.md5.asc
None
KNOPPIX_V5.0.1DVD-2006-06-01-DE.iso
None
KNOPPIX_V5.0.1DVD-2006-06-01-DE.iso.md5
None
KNOPPIX_V5.0.1DVD-2006-06-01-DE.iso.md5.asc
None
KNOPPIX_V5.0.1DVD-2006-06-01-EN.iso
None
KNOPPIX_V5.0.1DVD-2006-06-01-EN.iso.md5
None
KNOPPIX_V5.0.1DVD-2006-06-01-EN.iso.md5.asc
new  good redy wait old1 old2 old3 ntry conn cing busy all
   0    0    0    0    0    0    0    0    2    0    0    2 all source managers
(3 by UID) (1 ROQ)
   0    0    0    0    0    0    0    0    2    0    0    2     ready  2 active

  1. need sources

   0    0    0    0    0    0    0    0    0    0    0    0     throttled ready
   0    0    0    0    0    0    0    0    1    0    0    1     indirect
   0    0    0    0    0    0    0  600    0    0    0      period
Connecting Sources: 0 entries
Next Direct Sources: 0 entries
Next Indirect Sources: 0 entries
snap-----------------------------------

What should I do next? Do you need remote access to my machine?

cheers
  Kevin

spiralvoice <spiralvoice>
Group administrator
Thu 13 Jul 2006 03:48:43 PM UTC, comment #6: 

Mail from OP:
Hi!

I've made some progress: The error does not occur with my self compiled 2.7.2,
but the debian binary 2.7.3 is affected. In order to verify this is no debian
problem, I am now compiling 2.7.3 myself. My next test after that will be
logging verbosity -> all.

rgds
  Kevin

spiralvoice <spiralvoice>
Group administrator
Tue 11 Jul 2006 08:36:03 AM UTC, comment #5: 

On overview of 3rd-party libs:
since 2.5.30.10: libgd
since 2.6.1: libbz2
since 2.6.5: eMule SUI with CryptoPP
since 2.7.6: libmagic, GNU file

MLDonkey freezes even before libmagic was introduced, 2.6.5 with
all the other stuff works, so 3rd-party libs seem not to be the problem.
Between 2.6.6 and 2.7.2 were several changes which could be
responsible for the bug but I am not able to give a recommendation
which version to try next, unfortunately you have to give all of
them a try.

Another question about logging, if you set option verbosity to "all"
which are the last messages printed in log_file before MLDonkey freezes? Please try this with 2.7.7 because logging received some
updates in the latest versions.

spiralvoice <spiralvoice>
Group administrator
Tue 11 Jul 2006 08:27:36 AM UTC, comment #4: 

Mail fom OP:

Hi Spiralvoice!

Although I have only little time and that compiling takes ages, here are some
more results:

2.7.3-2 BUG
2.5.28 O.K.
2.7.7-2 BUG
2.6.4 = O.K.
2.6.5 = O.K.
2.7.7 with options disabled = BUG
2.7.3 = BUG

So the bug seems to have been introduced between 2.6.5 and 2.7.3. It seems to me
like it has nothing to do with the compile options.

Thank you for staying at it -- I hope it will take not much longer. My NSLU2 is
currently compiling 2.7.2. If you have any better idea than trying each single
version down to 2.6.6, please let me know so I save time!

krgds
  Kevin

spiralvoice <spiralvoice>
Group administrator
Sun 09 Jul 2006 09:31:36 PM UTC, comment #3: 

Try 2.7.7 with disabled options first, then from 2.7.3 backwards

spiralvoice <spiralvoice>
Group administrator
Sun 09 Jul 2006 09:30:35 PM UTC, comment #2: 

Mail from OP:

Hi Spiralvoice!

I've tested 2.6.4 and 2.6.5. Since my NSLU2 is quite slow (32MB RAM) I chose the
precompiled 2.6.4-1 from
<http://snapshot.debian.net/archive/2005/10/15/debian/pool/main/m/mldonkey/>
which runs fine. I then compiled 2.6.5 from
<http://download.savannah.gnu.org/releases/mldonkey/mldonkey-2.6.5.tar.bz2>
which also seems to run fine. So the error begins at some version between 2.6.5
and 2.7.3.

Do you have any other good guess which version to compile next? I would like to
avoid trying too many versions... Or is it best to try 2.7.7 with options
disabled next?

brgds
  Kevin

spiralvoice <spiralvoice>
Group administrator
Fri 07 Jul 2006 09:09:37 PM UTC, comment #1: 

I do not have access to ARM hardware, so you are on your own;-)
You are able to compile MLDonkey from sources yourself, right?

Could you please compile MLDonkey 2.6.4 and 2.6.5 and compare both
versions?
Emule Secure Ident was implemented in 2.6.5, its 3rd-party c++ source code and I do not know if it works stable on ARM hardware.
You could also try each version after 2.5.28 to see when the
bug was introduced, this means 2.5.28.1, 2.5.30, 2.5.30.1 and so
on, do not try 2.5.29. Grab the sourcecode with cvs -r or from
the files section on Savannah: https://savannah.nongnu.org/files/?group=mldonkey, for later
versions the sourcecode is on Sourceforge.

Another thing you can test with current 2.7.7 is to compile MLDonkey
with these settings:
./configure --disable-donkeysui --disable-gd --disable-bzip2 --disable-magic

If this core works enable the options step-by-step to see which
option produces the bug.
Its my first guess at the cause of this bug, especially when you
say 2.5.28 is not affected.

spiralvoice <spiralvoice>
Group administrator
Thu 06 Jul 2006 06:21:56 AM UTC, original submission:  

THIS ITEM WAS REASSIGNED TO PATCH #5752.

Hi!
The arm version of mldonkey-server reliably hangs a few minutes after I start it. This applies to the original debian testing version, as well as 2.7.3-2 sources manually compiled on sarge. It does not apply to version 2.5.28.

When it hangs, its CPU usage stays at 0% and it fails to respond to any connection attempts to ports 4000 and 4080 from localhost as well as the net.

I can easliy wake it up again by doing an "strace -p pid". This pid is not $((cat /var/run/mldonkey/mlnet.pid)) but rather the pid of the parent mlnet process. In the following pstree example this would be 14062.

|-mldonkey_server(14057)-+-logger(14064)
|                        |-logger(14066)
|                        `-mlnet(14062)---mlnet(14067)-+-mlnet(14068)
|                                                      `-mlnet(14069)

Once strace starts I can easily stop it again and mldonkey will run for another few minutes, until this situation occurs again.

I would be happy to deliver any further troubleshooting information to help you take this challenge. It seems to occur only on my arm machine, not i386. I'm not sure if it applies to all arms or only my Linksys NSLU2.

(See: http://www.cyrius.com/debian/nslu2/index.html)

You can get more information about the Debian bug report here:
http://bugs.debian.org/370107


Samuel Mimram <smimram>

 

(Note: upload size limit is set to 16384 kB, after insertion of the required escape characters.)

No files currently attached

 

Depends on the following items: None found

Items that depend on this one: None found

 

Carbon-Copy List
  • -email is unavailable- added by spiralvoice (Posted a comment)
  • -email is unavailable- added by smimram
  •  

    There are 0 votes so far. Votes easily highlight which items people would like to see resolved in priority, independently of the priority of the item set by tracker managers.

     

    Follow 17 latest changes.

    Date Changed by Updated Field Previous Value => Replaced by
    2007-02-18 spiralvoice Reassign Itemmldonkey, a multi-networks file-sharing client, bug #17041 mldonkey, a multi-networks file-sharing client, patch #5752
    2007-02-18 spiralvoice Attached File- Added arm.patch, #12004
    2007-02-18 spiralvoice Attached File#10357 Removed
    2006-07-21 spiralvoice Severity5 - Blocker 4 - Important
    2006-07-20 spiralvoice StatusIn Progress Postponed
    2006-07-20 spiralvoice Attached File#10371 Removed
    2006-07-19 spiralvoice Release2.7.3 2.7.7
    2006-07-18 spiralvoice Attached File- Added arm277.patch, #10371
    2006-07-15 spiralvoice Severity3 - Normal 5 - Blocker
        StatusNone In Progress
        ReleaseNone 2.7.3
        CPU typeNone ARM
    2006-07-15 spiralvoice Attached File- Added arm.patch, #10357
    2006-07-07 smimram Carbon-Copy- Added -email is unavailable-
    2006-07-07 smimram Carbon-CopyRemoved kevinprice<price.kevin@gmail.com> -
    2006-07-06 smimram Carbon-Copy- Added kevinprice<price.kevin@gmail.com>
    2006-07-06 smimram Carbon-Copy- Added -email is unavailable-

    Back to the top

    Powered by Savane 3.13-f8d8.
    Corresponding source code