bugmldonkey, a multi-networks file-sharing client - Bugs: bug #17933, After a few hours of core uptime,...

 
 

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

bug #17933: After a few hours of core uptime, it eats all CPU

Submitter:  Mathias Homann <lemmy98>
Submitted:  Fri 06 Oct 2006 07:55:35 PM UTC
   
 
Category:  Core Severity:  3 - Normal
Item Group:  Segmentation Fault Status:  Works For Me
Assigned to:  None Open/Closed:  Closed
Release:  2.8.1 Release:  2.8.1
Operating System:  Linux Binaries Origin:  CVS / Self compiled
CPU type:  Intel x86

Jump to the original submission

Wed 08 Nov 2006 03:41:41 PM UTC, comment #22: 

since the broken mlnet crashed inevitably when finishing a download, and the 2.8.1-cvs i've built earlier today has completed several so far, without crashing, i guess its really about suse breaking ocaml.

Mathias Homann <lemmy98>
Wed 08 Nov 2006 12:55:25 PM UTC, comment #21: 

I've built the cvs snapshot from http://www.mldonkey.org/forum/viewtopic.php?p=41793#41793 with no suse-provided ocaml rpm installed. let's see how it works out.

Mathias Homann <lemmy98>
Tue 07 Nov 2006 07:20:09 PM UTC, comment #20: 

Suse is known to have very buggy Ocaml packages:
http://www.mldonkey.org/forum/viewtopic.php?p=40997#40997

spiralvoice <spiralvoice>
Group administrator
Tue 07 Nov 2006 06:55:30 PM UTC, comment #19: 

Whenever I manage to find the time, I'll try the same with 2.8.1, and if that works out as well, I'll consider this case closeable...

Mathias Homann <lemmy98>
Tue 07 Nov 2006 06:54:36 PM UTC, comment #18: 

I dont know what exactly the problem was, but i've built a 2.7.7 core WITHOUT the suse-supplied ocaml rpm installed (so the makefile pulled ocaml from source and used it how it wanted it), and i've been running that core since 9 days without any trouble so far... I blame suse.

Mathias Homann <lemmy98>
Tue 07 Nov 2006 06:41:13 PM UTC, comment #17: 

Please re-compile with lesser CFLAGS

spiralvoice <spiralvoice>
Group administrator
Sun 29 Oct 2006 09:20:05 AM UTC, comment #16: 

I've upgraded my box to suse 10.0, and rebuilt mldonkey, and now ti doesnt hang and eat all cpu; instead it terminates between moving the completed download to its target folder, and writing to bittorrent.ini.

when terminating, it even breaks some other config files, namely downloads.ini, so that i have to grab some of the backups to get it back to work.

here's the buildinfo and runinfo output:

        --Buildinfo--
Version:         MLNet Multi-Network p2p client version 2.8.1
Networks:        Global Shares  FileTP  BitTorrent  Donkey (SUI)
Ocaml version:   3.09.3 - C compiler version: 4.0.2 - C++ compiler version: 4.0.2
Build on:        Linux i686 2.6.13-15.12-default (little endian) with glibc 2.3.5
Configure args:   '--prefix=/usr' '--disable-rpath' '--enable-magic' '--enable-gd' '--enable-batch' '--enable-gui' '--disable-gnutella' '--disable-gnutella2' '--disable-fasttrack' 'CC=gcc' 'CFLAGS=-O2 -g -march=i686 -mtune=i686 -fmessage-length=0 -D_FORTIFY_SOURCE=2 -W -Wall -Wno-unused -Wstrict-prototypes -Wpointer-arith'
Features:        threads zlib-1.2.3 bzip2-1.0.3 gd(jpg/png-1.2.8) iconv magic(active) no-check-bounds

runinfo

        --Runinfo--
User:            admin (PW Protected) - uptime: 7m 24s
Enabled nets:     Donkey Overnet BitTorrent FileTP
Server usage:    enabled
Geoip:           enabled, GeoLite data created by MaxMind, available from http://maxmind.com/
IP blocking:     local: 0 ranges - web: 110556 ranges
Libmagic:        file-type recognition database present
System info:     Linux celebrimbor 2.6.13-15.12-default #1 Thu Aug 24 11:23:58 UTC 2006 i686
                 language: DE - locale: UTF-8 - UTC offset: +0100
                 max_string_length: 16777211 - word_size: 32 - max_array_length: 4194303
                 max file descriptors: 1024 - max useable file size: 2^63-1 bits (do the maths ;-p)


Mathias Homann <lemmy98>
Thu 26 Oct 2006 06:51:26 PM UTC, comment #15: 

i'm starting to get the very same problem with my old 2.7.5 binary as well... the last few lines in the "verbosity" file before the segfaults start are something about "unknown clients". The fast things mlnet successfully does is movinga completed download into the target directory, and moving the .torrent file from torrents/downloads to torrents/seeded.
it does NOT remove the entry from files.ini.

Mathias Homann <lemmy98>
Mon 09 Oct 2006 06:12:29 PM UTC, comment #14: 

config_files_space.tmp is a normal file and stays behind when the core crashes.

You could also try to re-compile MLDonkey with
./configure --enable-minimum (maybe --enable-bittorrent)
to rule out disturbances by non-required features.

spiralvoice <spiralvoice>
Group administrator
Sun 08 Oct 2006 09:42:10 PM UTC, comment #13: 

another thing:
after 2.8.1 dieing, i find a 11meg sized file full of zeroes in its directory, named config_files_space.tmp

Mathias Homann <lemmy98>
Sun 08 Oct 2006 09:28:56 PM UTC, comment #12: 

same from the broken 2.8.1:

lemmy@celebrimbor:~/mldonkey-distrib-2.8.1> telnet localhost 4000
Trying ::1...
telnet: connect to address ::1: Connection refused
Trying 127.0.0.1...
Connected to localhost.
Escape character is '^]'.
Welcome to MLDonkey 2.8.1
Welcome on mldonkey command-line

Use ? for help

auth * ***
Full access enabled
diskinfo

        --Diskinfo--
Directory                          |Type                         |    used|    free|%free|Filesystem
-----------------------------------+-----------------------------+--------+--------+-----+----------
/users/lemmy/mldonkey-distrib-2.8.1|core/ini files               |  54.51G|   7.57G|  12%|reiserfs
/export/intranet/p2p/.mltemp       |temp/downloading             |   1.11G|   7.23G|  86%|ext2/3
/export/intranet/p2p/directories   |shared (incoming_directories)|   1.11G|   7.23G|  86%|ext2/3
/export/intranet/p2p/files         |shared (incoming_files)      |   1.11G|   7.23G|  86%|ext2/3
/export/intranet/p2p/shared        |shared (all_files)           |   1.11G|   7.23G|  86%|ext2/3

Mathias Homann <lemmy98>
Sun 08 Oct 2006 09:26:33 PM UTC, comment #11: 

diskinfo output taken with a 2.7.5:

mathias@pippin:~> telnet celebrimbor 4000
Trying 192.168.238.1...
Connected to celebrimbor.
Escape character is '^]'.
Welcome to MLDonkey 2.7.5
Welcome on mldonkey command-line

Use ? for help

auth * ***
Full access enabled
diskinfo

        --Diskinfo--
Directory                          |Type                         |    used|    free|%free|Filesystem
-----------------------------------+-----------------------------+--------+--------+-----+----------
/users/lemmy/mldonkey-distrib-2.7.5|core/ini files               |  54.52G|   7.56G|  12%|reiserfs
/export/intranet/p2p/.mltemp       |temp/downloading             |   1.11G|   7.23G|  86%|ext2/3
/export/intranet/p2p/directories   |shared (incoming_directories)|   1.11G|   7.23G|  86%|ext2/3
/export/intranet/p2p/files         |shared (incoming_files)      |   1.11G|   7.23G|  86%|ext2/3
/export/intranet/p2p/shared        |shared (all_files)           |   1.11G|   7.23G|  86%|ext2/3


Mathias Homann <lemmy98>
Sun 08 Oct 2006 06:14:29 PM UTC, comment #10: 

If you manage to bring MLDonkey into working state again
please post output of command "diskinfo".

You could also try to re-compile MLDonkey with
./configure --enable-minimum (maybe --enable-bittorrent)
to rule out disturbances by non-required features.

spiralvoice <spiralvoice>
Group administrator
Sun 08 Oct 2006 04:00:52 PM UTC, comment #9: 

I've restarted mlnet, and the completed download is still listed in there, with 99.8% completed.

mlnet is not responding on telnet, and is showing the same SIGSEGV strace output.


Mathias Homann <lemmy98>
Sun 08 Oct 2006 03:56:48 PM UTC, comment #8: 

there's nothing worthwhile in the log file:

2006/10/08 16:50:22 [dCon] Telnet connection from 192.168.238.1
2006/10/08 16:50:22 [dCon] Authenticated user: admin
2006/10/08 16:55:17 [dCon] Telnet connection from 192.168.238.1
2006/10/08 16:55:18 [dCon] Authenticated user: admin
2006/10/08 17:00:33 [dCon] Telnet connection from 192.168.238.1
2006/10/08 17:00:33 [dCon] Authenticated user: admin
2006/10/08 17:01:07 [cCO] Options correctly saved
2006/10/08 17:05:18 [dCon] Telnet connection from 192.168.238.1
2006/10/08 17:05:18 [dCon] Authenticated user: admin
mlnet.log lines 544-587/587 (END)    


those telnet connections are from a script that gets the throughput for munin.
at 17:07 mlnet completed a download; it's been hanging since then.
strace output:
--- SIGSEGV (Segmentation fault) ---      

repeated all over.

from directory timestamps, it looks like the last thing it did was a write access to the web_infos folder, or the moving of the torrent file from torrents/downloads to torrents/incoming.

Mathias Homann <lemmy98>
Sat 07 Oct 2006 08:03:35 PM UTC, comment #7: 

set log_file mlnet.log

and try again

spiralvoice <spiralvoice>
Group administrator
Sat 07 Oct 2006 07:26:26 PM UTC, comment #6: 

suse 8.0 right now...
lemmy@celebrimbor:~/mldonkey-distrib-2.8.1> ls *log
ls: *log: Datei oder Verzeichnis nicht gefunden

no log file...

Mathias Homann <lemmy98>
Sat 07 Oct 2006 06:57:37 PM UTC, comment #5: 

C compiler version: 2.95.3 <- which distro do you use? Eisfair?

How do the last lines of mlnet.log look like?

spiralvoice <spiralvoice>
Group administrator
Sat 07 Oct 2006 06:17:49 PM UTC, comment #4: 

re #1:

i tried to do that.
since it hangs in an endless loop and has to be killed -9, it doesnt produce profiling data...

Mathias Homann <lemmy98>
Sat 07 Oct 2006 05:56:50 PM UTC, comment #3: 

re #1:

since it does not terminate, and the bug is not really reproduceable except for the fact that it does surely hit after a while, profiling as described in the wiki would produce huge files with no real value in my opinion...
But i'll try anyways.

re #2:
here's the info:

 buildinfo

        --Buildinfo--
Version:         MLNet Multi-Network p2p client version 2.8.1
Networks:        Global Shares  Gnutella  G2  Fasttrack  FileTP  BitTorrent  Donkey (SUI)
Ocaml version:   3.08.3 - C compiler version: 2.95.3 - C++ compiler version: 2.95.3
Build on:        Linux i686 2.4.18-4GB (little endian) with glibc 2.2.5
Configure args:   '--disable-gui'
Features:        threads zlib-1.1.3 bzip2-1.0.2 gd(jpg/png-1.0.12) iconv no-magic no-check-bounds

runinfo

        --Runinfo--
User:            admin (PW Protected) - uptime: 0m 28s
Enabled nets:     Donkey Overnet BitTorrent FileTP
Server usage:    enabled
Geoip:           enabled, GeoLite data created by MaxMind, available from http://maxmind.com/
IP blocking:     local: 0 ranges - web: 107735 ranges
System info:     Linux celebrimbor 2.4.18-4GB #1 Wed May 11 23:24:34 CEST 2005 i686
                 language: JA - locale: UTF-8 - UTC offset: +0200
                 max_string_length: 16777211 - word_size: 32 - max_array_length: 4194303
                 max file descriptors: 1024 - max useable file size: 2^63-1 bits (do the maths ;-p)

Mathias Homann <lemmy98>
Sat 07 Oct 2006 03:40:26 PM UTC, comment #2: 

Please post buildinfo and runinfo output.
Here MLDonkey has uptimes of 1 month and more without the problem.

spiralvoice <spiralvoice>
Group administrator
Sat 07 Oct 2006 11:35:22 AM UTC, comment #1: 

please follow profiling instructions on wiki and return with info where cpu is going.

Joakim Verona <jave>
Group Member
Fri 06 Oct 2006 07:55:35 PM UTC, original submission:  

After a few hours of uptime (maybe related to finishing a download, it seems to start at that moment), mlnet hogs all cpu, stops responding on port 4000, and a strace -f -p $PID_OF_MLNET shows nothing but this:

--- SIGSEGV (Segmentation fault) ---
--- SIGSEGV (Segmentation fault) ---
--- SIGSEGV (Segmentation fault) ---
--- SIGSEGV (Segmentation fault) ---
--- SIGSEGV (Segmentation fault) ---
--- SIGSEGV (Segmentation fault) ---
--- SIGSEGV (Segmentation fault) ---
--- SIGSEGV (Segmentation fault) ---
--- SIGSEGV (Segmentation fault) ---
--- SIGSEGV (Segmentation fault) ---
--- SIGSEGV (Segmentation fault) ---
--- SIGSEGV (Segmentation fault) ---
--- SIGSEGV (Segmentation fault) ---
--- SIGSEGV (Segmentation fault) ---
--- SIGSEGV (Segmentation fault) ---
--- SIGSEGV (Segmentation fault) ---
--- SIGSEGV (Segmentation fault) ---
--- SIGSEGV (Segmentation fault) ---
--- SIGSEGV (Segmentation fault) ---
--- SIGSEGV (Segmentation fault) ---
--- SIGSEGV (Segmentation fault) ---
--- SIGSEGV (Segmentation fault) ---
--- SIGSEGV (Segmentation fault) ---
(repeat forever)

Mathias Homann <lemmy98>

 

(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

 

CC list is empty

 

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 2 latest changes.

Date Changed by Updated Field Previous Value => Replaced by
2006-11-07 spiralvoice StatusNone Works For Me
    Open/ClosedOpen Closed

Back to the top

Powered by Savane 3.13-02a9.
Corresponding source code