Amule client for wl500gp

[halleck]

Both and neither. Your problem seems to be with _llseek. offset_high values are unreasonable high. I suspect amule! Read this excerpt:

I can confirm that amuled 2.1.3 is crashing when downloading the files
with "Bad system call". Everything is fine when amuled is just seeding.
Maybe we will have to re-compile from the CVS.

[capes]

can't login via the amuleweb, i creat the remote.conf file, but can't longin.
any detail step for this funtion?

it doesn't generate the remote.conf file in /opt/share/amule/.aMule/ after i run amuleweb -w, Where is that file?

[capes]

it doesn't generate the remote.conf file in /opt/share/amule/.aMule/ after i run amuleweb -w, Where is that file?

first, create remote.conf file manually, and then edit the /opt/etc/init.d/S57amuled

...........
start() {
        set backup=$HOME
        echo "Starting amuled..."
        echo "Configuration's setting are located in /opt/share/amule"
        export HOME=/opt/share/amule
        /opt/bin/amuled -f
        /opt/bin/amuleweb --config-file=/opt/share/amule/.aMule/remote.conf --quiet &
        export HOME=$backup
        echo "done"
}
..........

[dimonb]

this script restart amuled if it crashes.
(it does not guard amuleweb)

while true; do

PID=`ps | grep amuled | grep -v grep`

if ! test -z "$PID"; then
        PID=`su amule -c "amulecmd -f ~/.aMule/amule.conf -c status" | grep "Succeeded! Connection established"`
fi

if test -z "$PID"; then
        #running problems, restart
        killall amuled 2>/dev/null
        killall amuleweb 2>/dev/null
        sleep 1
        killall -9 amuled 2>/dev/null
        killall -9 amuleweb 2>/dev/null
        sleep 1
        su amule -c"amuled -f >/dev/null"
fi

sleep 5
done

this script could be run with command like /opt/sbin/amule_check & in post-mount script

PS. Please, note script does not uses standard S57amuled to run amule, because running amule as a root is not a good idea IMHO.

[halleck]

this script restart amuled if it crashes.
(it does not guard amuleweb)

Well, restarting the daemon is not a proper solution - you may lose all
of your UL credits etc. Will try to look at re-compiling the amuled from
the latest CVS sources when my router comes back from the service.

[dimonb]

Well, restarting the daemon is not a proper solution - you may lose all
of your UL credits etc. Will try to look at re-compiling the amuled from
the latest CVS sources when my router comes back from the service.

This script restart daemon only if it's crashes or does not responding. Crashing of amule is terrible fact, but it's happens :-(((

it would be better to develop giFT plugin to work with eDonkey networks...

[al37919]

It seems that the large file crash problem is solved, thanks to the clue from dimonb. It is specific for uclibc and to some extent mipsel. I already posted detailed report in the Russian thread:

http://www.wl500g.info/showthread.php?p=51681#post51681

It happens because of buggy (or feature-rich?) setrlimit of uclibc. Maximum value of RLIMIT_FSIZE is RLIM_INFINITY which is defined as either (~0UL), or (~0UL >> 1) (latter includes also mips). setrlimit because of unknown idea of developer makes once again >> 1 . This is where 1Gb comes from.

If I execute 'strace amuled -f 2> log.log' I get:

getrlimit(RLIMIT_DATA, {rlim_cur=RLIM_INFINITY, rlim_max=RLIM_INFINITY}) = 0
setrlimit(RLIMIT_DATA, {rlim_cur=1073741823, rlim_max=1073741823}) = 0
getrlimit(RLIMIT_FSIZE, {rlim_cur=RLIM_INFINITY, rlim_max=RLIM_INFINITY}) = 0
setrlimit(RLIMIT_FSIZE, {rlim_cur=1073741823, rlim_max=1073741823}) = 0
getrlimit(RLIMIT_NOFILE, {rlim_cur=1024, rlim_max=1024}) = 0
setrlimit(RLIMIT_NOFILE, {rlim_cur=1024, rlim_max=1024}) = 0
getrlimit(RLIMIT_RSS, {rlim_cur=7401488, rlim_max=4735824}) = 0
setrlimit(RLIMIT_RSS, {rlim_cur=720283552, rlim_max=2147449912

After patching strace gives:

getrlimit(RLIMIT_DATA, {rlim_cur=RLIM_INFINITY, rlim_max=RLIM_INFINITY}) = 0
getrlimit(RLIMIT_FSIZE, {rlim_cur=RLIM_INFINITY, rlim_max=RLIM_INFINITY}) = 0
getrlimit(RLIMIT_NOFILE, {rlim_cur=1024, rlim_max=1024}) = 0
getrlimit(RLIMIT_RSS, {rlim_cur=RLIM_INFINITY, rlim_max=RLIM_INFINITY}) = 0

After patching it seems that the problem of maximum file size waived at all (at least till 4Gb size, as larger files are not supported by aMule at all):

_llseek(13, 4205578240, [4205578240], SEEK_SET) = 0
write(13, "\210\302\2568\10\25\322\16\225\350!\372\200OQ\6x\326\362"..., 10240) = 10240
_llseek(13, 4205598720, [4205578240], SEEK_SET) = 0
write(13, "\216\35{\312\347\305\271\247\243{\206\202\357\254\311`"..., 10240) = 10240
_llseek(13, 4205608960, [4205608960], SEEK_SET) = 0
write(13, "\36\250\2732;\354C!\356\232\376\352\365\202\177,~h\233"..., 10240) = 10240

2oleo or someone involved in development. Could you please "peer review" this stuff and if you find it sane submit in a right place. It's my first development under linux (which consists of a double slash ).

I suggest also to change S57amuled again to something like this:

#!/bin/sh
PATH=/sbin:/bin:/usr/bin:/usr/sbin:/opt/bin:/opt/sbin

#It is not safe to run amuled by the root
#Before the first run you have to create user p2p or whichever you like
USER=p2p
DIR=/opt/home/p2p
DAEMON=/opt/bin/amuled
NAME=amuled
NAME2=amuleweb

start() {
  echo "Configuration's setting are located in $DIR"
  echo -n "Starting $NAME... "
  su $USER -c "export HOME=$DIR && $DAEMON -f"
  echo "done"
}

stop() {
  echo -n "Shutting down $NAME... "
  killall $NAME
  killall $NAME2
  echo "done"
}

case "$1" in
	start)
		start
		;;
	stop)
		stop
		;;
	restart)
		stop
		sleep 1
		start
		;;
	*)
		echo "Usage: $0 (start|stop|restart)"
		exit 1
		;;
esac

I believe we have to enforce the user to chose more safe way.

[oleo]

Could you post a patch to amule with "diff -u old new" here. I have also upgraded uclibc to 0.9.28.3 and waiting to upgrade the feeds as all of them should be rebuilded.

[al37919]

OK, now I hope I did the patch correctly.

[oleo]

Patch added with http://trac.nslu2-linux.org/optware/changeset/5903

Not very clever though. We could also go undefining HAVE_ or setting

#ifndef __UCLIBC__

#endif 

to make it universal. But until then, ...

[al37919]

Even better approach would be to fix uclibc's setrlimit to make it arch dependent instead of weird >>1 , as any application in optware/oleg feed using this function has this problem.

However, at the moment this is not the biggest problem. After your compilation the issue is not solved. The patch is applied, because there are no calls to setrlimit, however, getrlimit returns 700Mb instead of RLIM_INFINITY.

getrlimit(RLIMIT_DATA, {rlim_cur=716175296, rlim_max=720250752}) = 0
getrlimit(RLIMIT_FSIZE, {rlim_cur=716175296, rlim_max=720250752}) = 0
getrlimit(RLIMIT_NOFILE, {rlim_cur=716175296, rlim_max=720250752}) = 0
getrlimit(RLIMIT_RSS, {rlim_cur=7401488, rlim_max=9442296}) = 0

So, in the new build file-size-limit is lower than before. This behavior is not related to my patch, but might be due to the changes in the uclibc library, or differences in the build system, (or even in the execution environment?). The second is most likely. The size of my amuled binary is 3723Mb, the size of you binary is 3243Mb.

Here is my build system (created according to http://www.nslu2-linux.org/wiki/Optw...ckageToOptware ):
openSuSE 10.2
mkdir /opt/optware
cd /opt/optware
svn co http://svn.nslu2-linux.org/svnroot/optware/trunk optware
edit /opt/optware/Makefile : OPTWARE_TARGET ?= oleg
make directories ipkg-utils
make

After compiling toolchain and creation of about 10 packs (including amule-2.1.3-6.ipk) I stopped it. Then I edited /opt/optware/builds/amule/src/amule.cpp

I was compiling it while being in /opt/optware with make amule

What is wrong?

[oleo]

Could you please verify if the updated uclibc library works without your patch. I am planning to merge uclibc-unstable branch back to trunk.
Here is what you do:

svn co http://svn.nslu2-linux.org/svnroot/optware/branches/uclibc-unstable unstable
cd unstable
export OPTWARE_TARGET=oleg
make directories 
make toolchain
make amule

You will need to replace all packages on your system with packages that you built. This means at least uclibc-opt, ipkg-opt, amule and dependent packages.

Newest toolchain uses gcc 4.1.2 and latest uclibc 0.9.28.3
I did not verify if limits work as espected, but I've created this branch to softland mediatomb exception handling with SJLJ.

If you will find that amule have the same symptmoms, then we will create patch for uclibc! Either by backport from 0.9.29 or from scratch.

[al37919]

OK, I hope I did what you've expected.
I compiled the unstable toolchain and packages as you wrote. Then I set up a spare flash-drive as /opt and started to install newly compiled packages as usually:
mkdir /opt/tmp
mkdir /opt/tmp/ipkg
ipkg.sh update
ipkg.sh install ipkg

after that i started to install locally compiled packages. I haven't succeed only with ipkg-opt, as ipkg was already installed

amule shows:

getrlimit(RLIMIT_DATA, {rlim_cur=RLIM_INFINITY, rlim_max=RLIM_INFINITY}) = 0
setrlimit(RLIMIT_DATA, {rlim_cur=704010*1024, rlim_max=4674376}) = 0
getrlimit(RLIMIT_FSIZE, {rlim_cur=RLIM_INFINITY, rlim_max=RLIM_INFINITY}) = 0
setrlimit(RLIMIT_FSIZE, {rlim_cur=704010*1024, rlim_max=4674376}) = 0
getrlimit(RLIMIT_NOFILE, {rlim_cur=RLIM_INFINITY, rlim_max=RLIM_INFINITY}) = 0
setrlimit(RLIMIT_NOFILE, {rlim_cur=1024, rlim_max=1024}) = 0
getrlimit(RLIMIT_RSS, {rlim_cur=2147450584, rlim_max=4674512}) = 0
setrlimit(RLIMIT_RSS, {rlim_cur=704010*1024, rlim_max=2147449144}) = 0

as a response to

static void UnlimitResource(RLIMIT_RESOURCE resType)
{
#if defined(HAVE_GETRLIMIT) && defined(HAVE_SETRLIMIT)
	struct rlimit rl;
	getrlimit(resType, &rl);
	rl.rlim_cur = rl.rlim_max;
	setrlimit(resType, &rl);
#endif
}

Do you understand something? I can only say that while compiling on my computer I get correct initial RLIMITs in all cases. setrlimits works wrong in both cases.

[oleo]

To isolate the problem I suggest the folowing program rlimit.c

#include <sys/time.h>
#include <sys/resource.h>

static void UnlimitResource(int resType)
{
        struct rlimit rl;
        getrlimit(resType, &rl);
        rl.rlim_cur = rl.rlim_max;
        setrlimit(resType, &rl);
        getrlimit(resType, &rl);
}

int main()
{
        UnlimitResource(RLIMIT_DATA);
        UnlimitResource(RLIMIT_FSIZE);
        UnlimitResource(RLIMIT_NOFILE);
        UnlimitResource(RLIMIT_RSS);
} 

Compile it with buildroot-4.1.2-13 under router with
gcc rlimit.c and report its output of strace ./a.out.

When I've compiled it with buildroot-4.1.1-12 I've got the following result:

getrlimit(RLIMIT_DATA, {rlim_cur=RLIM_INFINITY, rlim_max=RLIM_INFINITY}) = 0
setrlimit(RLIMIT_DATA, {rlim_cur=1073741823, rlim_max=1073741823}) = 0
getrlimit(RLIMIT_DATA, {rlim_cur=1073741823, rlim_max=1073741823}) = 0
getrlimit(RLIMIT_FSIZE, {rlim_cur=RLIM_INFINITY, rlim_max=RLIM_INFINITY}) = 0
setrlimit(RLIMIT_FSIZE, {rlim_cur=1073741823, rlim_max=1073741823}) = 0
getrlimit(RLIMIT_FSIZE, {rlim_cur=1073741823, rlim_max=1073741823}) = 0
getrlimit(RLIMIT_NOFILE, {rlim_cur=1024, rlim_max=1024}) = 0
setrlimit(RLIMIT_NOFILE, {rlim_cur=1024, rlim_max=1024}) = 0
getrlimit(RLIMIT_NOFILE, {rlim_cur=1024, rlim_max=1024}) = 0
getrlimit(RLIMIT_RSS, {rlim_cur=RLIM_INFINITY, rlim_max=RLIM_INFINITY}) = 0
setrlimit(RLIMIT_RSS, {rlim_cur=1073741823, rlim_max=1073741823}) = 0
getrlimit(RLIMIT_RSS, {rlim_cur=1073741823, rlim_max=1073741823}) = 0
exit(0)                                 = ?

That shows filesize limit 1GB limit with current trunk oleg feed.

Please give gcc --version to verify that you have correct buildroot and uclibc.

So you need at least ipkg-opt, uclibc-opt, buildroot and strace from unstable branch to verify this.

[al37919]

OK, I did it. I have native-compiled it in the two cases (unstable and stable). The both indicate the same 1Gb problem as you shown and as I shown in my initial post. I'm confused only with uclibc-opt - 0.9.28-13, as I would expect to find here uclibc-opt - 0.9.28.3.

unstable:

root@wl500g,/opt/tmp> ipkg list_installed
...
buildroot - 4.1.2-13 -
libuclibc++ - 0.2.2-5 -
uclibc-opt - 0.9.28-13 -
...
root@wl500g,/opt/tmp> gcc --version
gcc (GCC) 4.1.2
...
root@wl500g,/opt/tmp> gcc rlimit.c
root@wl500g,/opt/tmp> strace ./a.out
...
getrlimit(RLIMIT_DATA, {rlim_cur=RLIM_INFINITY, rlim_max=RLIM_INFINITY}) = 0
setrlimit(RLIMIT_DATA, {rlim_cur=1073741823, rlim_max=1073741823}) = 0
getrlimit(RLIMIT_DATA, {rlim_cur=1073741823, rlim_max=1073741823}) = 0
getrlimit(RLIMIT_FSIZE, {rlim_cur=RLIM_INFINITY, rlim_max=RLIM_INFINITY}) = 0
setrlimit(RLIMIT_FSIZE, {rlim_cur=1073741823, rlim_max=1073741823}) = 0
getrlimit(RLIMIT_FSIZE, {rlim_cur=1073741823, rlim_max=1073741823}) = 0
getrlimit(RLIMIT_NOFILE, {rlim_cur=1024, rlim_max=1024}) = 0
setrlimit(RLIMIT_NOFILE, {rlim_cur=1024, rlim_max=1024}) = 0
getrlimit(RLIMIT_NOFILE, {rlim_cur=1024, rlim_max=1024}) = 0
getrlimit(RLIMIT_RSS, {rlim_cur=RLIM_INFINITY, rlim_max=RLIM_INFINITY}) = 0
setrlimit(RLIMIT_RSS, {rlim_cur=1073741823, rlim_max=1073741823}) = 0
getrlimit(RLIMIT_RSS, {rlim_cur=1073741823, rlim_max=1073741823}) = 0 

stable:

root@wl500g,/opt/tmp> ipkg list_installed | grep -e u[cC]libc
buildroot - 4.1.1-12 - uClibc compilation toolchain
libuclibc++ - 0.2.2-4 - C++ standard library designed for use in embedded systems
uclibc-opt - 0.9.28-12 - micro C library for embedded Linux systems
root@wl500g,/opt/tmp> gcc --version
gcc (GCC) 4.1.1
...
root@wl500g,/opt/tmp> gcc rlimit.c
root@wl500g,/opt/tmp> strace ./a.out
...
getrlimit(RLIMIT_DATA, {rlim_cur=RLIM_INFINITY, rlim_max=RLIM_INFINITY}) = 0
setrlimit(RLIMIT_DATA, {rlim_cur=1073741823, rlim_max=1073741823}) = 0
getrlimit(RLIMIT_DATA, {rlim_cur=1073741823, rlim_max=1073741823}) = 0
getrlimit(RLIMIT_FSIZE, {rlim_cur=RLIM_INFINITY, rlim_max=RLIM_INFINITY}) = 0
setrlimit(RLIMIT_FSIZE, {rlim_cur=1073741823, rlim_max=1073741823}) = 0
getrlimit(RLIMIT_FSIZE, {rlim_cur=1073741823, rlim_max=1073741823}) = 0
getrlimit(RLIMIT_NOFILE, {rlim_cur=1024, rlim_max=1024}) = 0
setrlimit(RLIMIT_NOFILE, {rlim_cur=1024, rlim_max=1024}) = 0
getrlimit(RLIMIT_NOFILE, {rlim_cur=1024, rlim_max=1024}) = 0
getrlimit(RLIMIT_RSS, {rlim_cur=RLIM_INFINITY, rlim_max=RLIM_INFINITY}) = 0
setrlimit(RLIMIT_RSS, {rlim_cur=1073741823, rlim_max=1073741823}) = 0
getrlimit(RLIMIT_RSS, {rlim_cur=1073741823, rlim_max=1073741823}) = 0