Pavol Kycina
|
mbuf leak (related to pppd)
|
Pavol Kycina
10/22/2009 11:53 AM
post40506
|
mbuf leak (related to pppd)
I have discovered an mbuf memory leak (recognizable by netstat -m).
Each invocation of pppd "consumes" one mbuf.
I have a small system (memory constrained) and it creates a problem for me.
Any ideas?
Thanks, Pavel Kycina
|
|
|
Sean Boudreau(deleted)
|
Re: mbuf leak (related to pppd)
|
Sean Boudreau(deleted)
10/23/2009 9:02 AM
post40600
|
Re: mbuf leak (related to pppd)
On Thu, Oct 22, 2009 at 11:53:57AM -0400, Pavol Kycina wrote:
> I have discovered an mbuf memory leak (recognizable by netstat -m).
>
> Each invocation of pppd "consumes" one mbuf.
>
> I have a small system (memory constrained) and it creates a problem for me.
I've reproduced this. I'll keep you posted...
-seanb
|
|
|
Sean Boudreau(deleted)
|
Re: mbuf leak (related to pppd)
|
Sean Boudreau(deleted)
10/24/2009 10:24 PM
post40649
|
Re: mbuf leak (related to pppd)
On Thu, Oct 22, 2009 at 11:53:57AM -0400, Pavol Kycina wrote:
> I have discovered an mbuf memory leak (recognizable by netstat -m).
>
> Each invocation of pppd "consumes" one mbuf.
>
> I have a small system (memory constrained) and it creates a problem for me.
>
> Any ideas?
Should be fixed now.
Regards,
-seanb
|
|
|
Pavol Kycina
|
Re: mbuf leak (related to pppd)
|
Pavol Kycina
10/26/2009 7:39 AM
post40661
|
Re: mbuf leak (related to pppd)
Thanks for good info.
As there is a message about maintenance of svn repository, would it be possible to send me built version of pppd?
Regards, PKY
> On Thu, Oct 22, 2009 at 11:53:57AM -0400, Pavol Kycina wrote:
> > I have discovered an mbuf memory leak (recognizable by netstat -m).
> >
> > Each invocation of pppd "consumes" one mbuf.
> >
> > I have a small system (memory constrained) and it creates a problem for me.
> >
> > Any ideas?
>
> Should be fixed now.
>
> Regards,
>
> -seanb
|
|
|
Sean Boudreau(deleted)
|
Re: mbuf leak (related to pppd)
|
Sean Boudreau(deleted)
10/26/2009 8:14 AM
post40662
|
Re: mbuf leak (related to pppd)
On Mon, Oct 26, 2009 at 07:39:24AM -0400, Pavol Kycina wrote:
> Thanks for good info.
>
> As there is a message about maintenance of svn repository, would it be possible to send me built version of pppd?
It's back up now. These changes are therein.
-seanb
|
|
|
Sean Boudreau(deleted)
|
Re: mbuf leak (related to pppd)
|
Sean Boudreau(deleted)
10/26/2009 8:20 AM
post40663
|
Re: mbuf leak (related to pppd)
> On Mon, Oct 26, 2009 at 07:39:24AM -0400, Pavol Kycina wrote:
> > Thanks for good info.
> >
> > As there is a message about maintenance of svn repository, would it be possible to send me built version of pppd?
>
> It's back up now. These changes are therein.
BTW the fix for this is in the stack, not pppd.
-seanb
|
|
|
Pavol Kycina
|
Re: mbuf leak (related to pppd)
|
Pavol Kycina
10/26/2009 9:50 AM
post40676
|
Re: mbuf leak (related to pppd)
It's probably back online, but I can't authorize to it. (though I am using the same password as to other repositories...
)
Could you please send me corrected io-pkt-v4?
Thanks, PKY
> > On Mon, Oct 26, 2009 at 07:39:24AM -0400, Pavol Kycina wrote:
> > > Thanks for good info.
> > >
> > > As there is a message about maintenance of svn repository, would it be
> possible to send me built version of pppd?
> >
> > It's back up now. These changes are therein.
>
> BTW the fix for this is in the stack, not pppd.
>
> -seanb
|
|
|
Sean Boudreau(deleted)
|
Re: mbuf leak (related to pppd)
|
Sean Boudreau(deleted)
10/26/2009 1:52 PM
post40706
|
Re: mbuf leak (related to pppd)
On Mon, Oct 26, 2009 at 09:50:22AM -0400, Pavol Kycina wrote:
> It's probably back online, but I can't authorize to it. (though I am using the same password as to other repositories.
..)
>
> Could you please send me corrected io-pkt-v4?
>
I checked with the powers that be and it should be up again.
Try checking it out to a new location rather than updating.
Regards,
-seanb
|
|
|
Pavol Kycina
|
Re: mbuf leak (related to pppd)
|
Pavol Kycina
10/27/2009 6:20 AM
post40742
|
Re: mbuf leak (related to pppd)
> On Mon, Oct 26, 2009 at 09:50:22AM -0400, Pavol Kycina wrote:
> > It's probably back online, but I can't authorize to it. (though I am using
> the same password as to other repositories...)
> >
> > Could you please send me corrected io-pkt-v4?
> >
>
> I checked with the powers that be and it should be up again.
> Try checking it out to a new location rather than updating.
>
> Regards,
>
> -seanb
Thanks, I have managed to check it out.
But I have found these two problems:
First (I managed to solve):
/tmp/qnx_core_net2/lib/io-pkt/lib/socket/net/getpeereid.c
there is no header for the function in this file.
Second (stopper for me):
make -j 1 -Cx86 -fMakefile
make[3]: Entering directory `/tmp/qnx_core_net2/lib/io-pkt/lib/dns/x86'
make -j 1 -Ca -fMakefile
make[4]: Entering directory `/tmp/qnx_core_net2/lib/io-pkt/lib/dns/x86/a'
make[4]: Nothing to be done for `first'.
make[4]: Leaving directory `/tmp/qnx_core_net2/lib/io-pkt/lib/dns/x86/a'
make[3]: Leaving directory `/tmp/qnx_core_net2/lib/io-pkt/lib/dns/x86'
make[2]: Leaving directory `/tmp/qnx_core_net2/lib/io-pkt/lib/dns'
make -j 1 -Cedit -fMakefile
make[2]: Entering directory `/tmp/qnx_core_net2/lib/io-pkt/lib/edit'
make -j 1 -Cx86 -fMakefile
make[3]: Entering directory `/tmp/qnx_core_net2/lib/io-pkt/lib/edit/x86'
make -j 1 -Ca -fMakefile
make[4]: Entering directory `/tmp/qnx_core_net2/lib/io-pkt/lib/edit/x86/a'
/usr/qnx641/host/qnx6/x86/usr/bin/qcc -Vgcc_ntox86 -c -Wc,-Wall -Wc,-Wno-parentheses
-O -DNDEBUG -I. -I/tmp/qnx_core_net2/lib/io-pkt/lib/edit -I/usr/qnx641/target/qnx6/u
sr/include -DVARIANT_a -DBUILDENV_qss /tmp/qnx_core_net2/lib/io-pkt/lib/edit/r
eadline.c
In file included from /tmp/qnx_core_net2/lib/io-pkt/lib/edit/el.h:105,
from /tmp/qnx_core_net2/lib/io-pkt/lib/edit/readline.c:60:
/tmp/qnx_core_net2/lib/io-pkt/lib/edit/map.h:59: error: expected specifier-qualifier-l
ist before 'el_func_t'
/tmp/qnx_core_net2/lib/io-pkt/lib/edit/map.h:73: error: expected declaration specifier
s or '...' before 'el_func_t'
/tmp/qnx_core_net2/lib/io-pkt/lib/edit/readline.c: In function 'rl_bind_key':
/tmp/qnx_core_net2/lib/io-pkt/lib/edit/readline.c:1568: error: 'ED_INSERT' undeclared
(first use in this function)
/tmp/qnx_core_net2/lib/io-pkt/lib/edit/readline.c:1568: error: (Each undeclared identi
fier is reported only once
/tmp/qnx_core_net2/lib/io-pkt/lib/edit/readline.c:1568: error: for each function it ap
pears in.)
cc: /usr/qnx641/host/qnx6/x86/usr/lib/gcc/i386-pc-nto-qnx6.4.0/4.3.3/cc1 error 1
make[4]: *** [readline.o] Error 1
make[4]: Leaving directory `/tmp/qnx_core_net2/lib/io-pkt/lib/edit/x86/a'
make[3]: *** [all] Error 2
make[3]: Leaving directory `/tmp/qnx_core_net2/lib/io-pkt/lib/edit/x86'
make[2]: *** [all] Error 2
make[2]: Leaving directory `/tmp/qnx_core_net2/lib/io-pkt/lib/edit'
make[1]: *** [all] Error 2
make[1]: Leaving directory `/tmp/qnx_core_net2/lib/io-pkt/lib'
make: *** [all] Error 2
Any idea what went wrong?
Thanks, PKY
|
|
|
Pavol Kycina
|
Re: mbuf leak (related to pppd)
|
Pavol Kycina
10/27/2009 6:32 AM
post40743
|
Re: mbuf leak (related to pppd)
The second problem was caused by not having executable permissions on makelist in that dir.
But soon after that in
/tmp/qnx_core_net2/lib/io-pkt/lib/io-pkt/crypto/dist2/openssl/ssl/ it was unable to compile bio_ssl.c, with many errors
in ssl.h
Any idea?
Thanks, PKY
> > On Mon, Oct 26, 2009 at 09:50:22AM -0400, Pavol Kycina wrote:
> > > It's probably back online, but I can't authorize to it. (though I am using
>
> > the same password as to other repositories...)
> > >
> > > Could you please send me corrected io-pkt-v4?
> > >
> >
> > I checked with the powers that be and it should be up again.
> > Try checking it out to a new location rather than updating.
> >
> > Regards,
> >
> > -seanb
>
>
> Thanks, I have managed to check it out.
>
> But I have found these two problems:
>
> First (I managed to solve):
> /tmp/qnx_core_net2/lib/io-pkt/lib/socket/net/getpeereid.c
> there is no header for the function in this file.
>
> Second (stopper for me):
>
> make -j 1 -Cx86 -fMakefile
> make[3]: Entering directory `/tmp/qnx_core_net2/lib/io-pkt/lib/dns/x86'
> make -j 1 -Ca -fMakefile
> make[4]: Entering directory `/tmp/qnx_core_net2/lib/io-pkt/lib/dns/x86/a'
> make[4]: Nothing to be done for `first'.
> make[4]: Leaving directory `/tmp/qnx_core_net2/lib/io-pkt/lib/dns/x86/a'
> make[3]: Leaving directory `/tmp/qnx_core_net2/lib/io-pkt/lib/dns/x86'
> make[2]: Leaving directory `/tmp/qnx_core_net2/lib/io-pkt/lib/dns'
> make -j 1 -Cedit -fMakefile
> make[2]: Entering directory `/tmp/qnx_core_net2/lib/io-pkt/lib/edit'
> make -j 1 -Cx86 -fMakefile
> make[3]: Entering directory `/tmp/qnx_core_net2/lib/io-pkt/lib/edit/x86'
> make -j 1 -Ca -fMakefile
> make[4]: Entering directory `/tmp/qnx_core_net2/lib/io-pkt/lib/edit/x86/a'
> /usr/qnx641/host/qnx6/x86/usr/bin/qcc -Vgcc_ntox86 -c -Wc,-Wall -Wc,-Wno-
> parentheses
> -O -DNDEBUG -I. -I/tmp/qnx_core_net2/lib/io-pkt/lib/edit -I/usr/qnx641/
> target/qnx6/u
> sr/include -DVARIANT_a -DBUILDENV_qss /tmp/qnx_core_net2/lib/io-pkt/
> lib/edit/r
> eadline.c
> In file included from /tmp/qnx_core_net2/lib/io-pkt/lib/edit/el.h:105,
> from /tmp/qnx_core_net2/lib/io-pkt/lib/edit/readline.c:60:
> /tmp/qnx_core_net2/lib/io-pkt/lib/edit/map.h:59: error: expected specifier-
> qualifier-l
> ist before 'el_func_t'
> /tmp/qnx_core_net2/lib/io-pkt/lib/edit/map.h:73: error: expected declaration
> specifier
> s or '...' before 'el_func_t'
> /tmp/qnx_core_net2/lib/io-pkt/lib/edit/readline.c: In function 'rl_bind_key':
> /tmp/qnx_core_net2/lib/io-pkt/lib/edit/readline.c:1568: error: 'ED_INSERT'
> undeclared
> (first use in this function)
> /tmp/qnx_core_net2/lib/io-pkt/lib/edit/readline.c:1568: error: (Each
> undeclared identi
> fier is reported only once
> /tmp/qnx_core_net2/lib/io-pkt/lib/edit/readline.c:1568: error: for each
> function it ap
> pears in.)
> cc: /usr/qnx641/host/qnx6/x86/usr/lib/gcc/i386-pc-nto-qnx6.4.0/4.3.3/cc1 error
> 1
> make[4]: *** [readline.o] Error 1
> make[4]: Leaving directory `/tmp/qnx_core_net2/lib/io-pkt/lib/edit/x86/a'
> make[3]: *** [all] Error 2
> make[3]: Leaving directory `/tmp/qnx_core_net2/lib/io-pkt/lib/edit/x86'
> make[2]: *** [all] Error 2
> make[2]: Leaving directory `/tmp/qnx_core_net2/lib/io-pkt/lib/edit'
> make[1]: *** [all] Error 2
> make[1]: Leaving directory `/tmp/qnx_core_net2/lib/io-pkt/lib'
> make: *** [all] Error 2
>
> Any idea what went wrong?
>
> Thanks, PKY
|
|
|
Sean Boudreau(deleted)
|
Re: mbuf leak (related to pppd)
|
Sean Boudreau(deleted)
10/27/2009 8:17 AM
post40760
|
Re: mbuf leak (related to pppd)
On Tue, Oct 27, 2009 at 06:32:47AM -0400, Pavol Kycina wrote:
> The second problem was caused by not having executable permissions on makelist in that dir.
>
> But soon after that in
> /tmp/qnx_core_net2/lib/io-pkt/lib/io-pkt/crypto/dist2/openssl/ssl/ it was unable to compile bio_ssl.c, with many
errors in ssl.h
>
> Any idea?
You're probably trying to compile for arm
with an old compiler. You can:
- exclude arm
# make EXCLUDE_CPULIST="arm"
# make CPULIST="x86 ppc"
- compile that particular file for arm with -O0
# cd lib/ssl/arm && make CCOPTS=-O0 bio_ssl.o
then continue with your previous lower level make
-seanb
|
|
|
Pavol Kycina
|
Re: mbuf leak (related to pppd)
|
Pavol Kycina
10/27/2009 10:57 AM
post40787
|
Re: mbuf leak (related to pppd)
Even the first time I was making it for x86 only,
make CPULIST="x86"
Now I have tried
make EXCLUDE_CPULIST="arm"
And the result is the same: not possible to compile bio_ssl.c
My qconfig says:
QNX SDP 6.4.1
Thanks, PKY
> On Tue, Oct 27, 2009 at 06:32:47AM -0400, Pavol Kycina wrote:
> > The second problem was caused by not having executable permissions on
> makelist in that dir.
> >
> > But soon after that in
> > /tmp/qnx_core_net2/lib/io-pkt/lib/io-pkt/crypto/dist2/openssl/ssl/ it was
> unable to compile bio_ssl.c, with many errors in ssl.h
> >
> > Any idea?
>
> You're probably trying to compile for arm
> with an old compiler. You can:
>
> - exclude arm
> # make EXCLUDE_CPULIST="arm"
> # make CPULIST="x86 ppc"
> - compile that particular file for arm with -O0
> # cd lib/ssl/arm && make CCOPTS=-O0 bio_ssl.o
> then continue with your previous lower level make
>
> -seanb
|
|
|
Sean Boudreau(deleted)
|
RE: mbuf leak (related to pppd)
|
Sean Boudreau(deleted)
10/27/2009 10:58 AM
post40788
|
RE: mbuf leak (related to pppd)
Please post the errors.
Thanks,
-seanb
-----Original Message-----
From: Pavol Kycina [mailto:community-noreply@qnx.com]
Sent: Tuesday, October 27, 2009 10:57 AM
To: general-networking
Subject: Re: mbuf leak (related to pppd)
Even the first time I was making it for x86 only,
make CPULIST="x86"
Now I have tried
make EXCLUDE_CPULIST="arm"
And the result is the same: not possible to compile bio_ssl.c
My qconfig says:
QNX SDP 6.4.1
|
|
|
Pavol Kycina
|
Re: RE: mbuf leak (related to pppd)
|
Pavol Kycina
10/27/2009 11:10 AM
post40791
|
Re: RE: mbuf leak (related to pppd)
Attached file is the output of:
make CPULIST=x86 2>/tmp/qnx_corenet.log_err
PKY
> Please post the errors.
>
> Thanks,
>
> -seanb
>
> -----Original Message-----
> From: Pavol Kycina [mailto:community-noreply@qnx.com]
> Sent: Tuesday, October 27, 2009 10:57 AM
> To: general-networking
> Subject: Re: mbuf leak (related to pppd)
>
> Even the first time I was making it for x86 only,
>
> make CPULIST="x86"
>
> Now I have tried
>
> make EXCLUDE_CPULIST="arm"
>
> And the result is the same: not possible to compile bio_ssl.c
>
> My qconfig says:
>
> QNX SDP 6.4.1
|
|
|
Sean Boudreau(deleted)
|
Re: RE: mbuf leak (related to pppd)
|
Sean Boudreau(deleted)
10/27/2009 11:17 AM
post40794
|
Re: RE: mbuf leak (related to pppd)
On Tue, Oct 27, 2009 at 11:10:02AM -0400, Pavol Kycina wrote:
> Attached file is the output of:
>
> make CPULIST=x86 2>/tmp/qnx_corenet.log_err
You need to do a 'make CPULIST=x86 install'
from the root. I believe this should be
solved if libcrypto is built beforehand
(which it should be by default) and installed
to your stage.
-seasnb
|
|
|
Pavol Kycina
|
Re: RE: mbuf leak (related to pppd)
|
Pavol Kycina
10/28/2009 2:53 AM
post40858
|
Re: RE: mbuf leak (related to pppd)
Thanks,
Now I have compiled executable.
But at the same time I have development system in "non defined" state, as make install didn't work on stageing area, but
on real system.
I have followed:
http://community.qnx.com/sf/wiki/do/viewPage/projects.core_os/wiki/BuildTheOSSource
this is the state of my system:
# pwd
/tmp/qnx_core_net2/lib/io-pkt
#
# set | grep QCONF
QCONF_OVERRIDE=/tmp/qnx_core_net2/lib/io-pkt/qconf-override.mk
#
# cat qconf_override.mk
USE_INSTALL_ROOT=1
INSTALL_ROOT_nto=/tmp/qnx_core_net2/lib/io-pkt/stage
#
# ls -D
. common dist external services sys
.. crypto dist2 lib stage utils
#
#
After make CPULIST=x86 hinstall and also make CPULIST=x86 install the stage directory is empty.
All cp commands invoked by make were copying to real system.
Any idea why?
Thanks, PKY
> On Tue, Oct 27, 2009 at 11:10:02AM -0400, Pavol Kycina wrote:
> > Attached file is the output of:
> >
> > make CPULIST=x86 2>/tmp/qnx_corenet.log_err
>
> You need to do a 'make CPULIST=x86 install'
> from the root. I believe this should be
> solved if libcrypto is built beforehand
> (which it should be by default) and installed
> to your stage.
>
> -seasnb
|
|
|
Sean Boudreau(deleted)
|
Re: RE: mbuf leak (related to pppd)
|
Sean Boudreau(deleted)
10/28/2009 9:40 AM
post40878
|
Re: RE: mbuf leak (related to pppd)
On Wed, Oct 28, 2009 at 02:53:07AM -0400, Pavol Kycina wrote:
> Thanks,
>
> Now I have compiled executable.
>
> But at the same time I have development system in "non defined" state, as make install didn't work on stageing area,
but on real system.
>
> I have followed:
> http://community.qnx.com/sf/wiki/do/viewPage/projects.core_os/wiki/BuildTheOSSource
>
> this is the state of my system:
>
> # pwd
> /tmp/qnx_core_net2/lib/io-pkt
> #
> # set | grep QCONF
> QCONF_OVERRIDE=/tmp/qnx_core_net2/lib/io-pkt/qconf-override.mk
> #
> # cat qconf_override.mk
You've exported 'qconf-override.mk' but setup 'qconf_override.mk'
('-' vs '_'). This is why you shouldn't 'make install' as root
(at least the first time). :)
-seanb
|
|
|
Pavol Kycina
|
Re: RE: mbuf leak (related to pppd)
|
Pavol Kycina
10/28/2009 9:49 AM
post40881
|
Re: RE: mbuf leak (related to pppd)
Thanks for all the assistance.
PKY
> On Wed, Oct 28, 2009 at 02:53:07AM -0400, Pavol Kycina wrote:
> > Thanks,
> >
> > Now I have compiled executable.
> >
> > But at the same time I have development system in "non defined" state, as
> make install didn't work on stageing area, but on real system.
> >
> > I have followed:
> > http://community.qnx.com/sf/wiki/do/viewPage/projects.core_os/wiki/
> BuildTheOSSource
> >
> > this is the state of my system:
> >
> > # pwd
> > /tmp/qnx_core_net2/lib/io-pkt
> > #
> > # set | grep QCONF
> > QCONF_OVERRIDE=/tmp/qnx_core_net2/lib/io-pkt/qconf-override.mk
> > #
> > # cat qconf_override.mk
>
> You've exported 'qconf-override.mk' but setup 'qconf_override.mk'
> ('-' vs '_'). This is why you shouldn't 'make install' as root
> (at least the first time). :)
>
>
> -seanb
|
|
|
Pavol Kycina
|
Re: RE: mbuf leak (related to pppd)
|
Pavol Kycina
10/28/2009 10:14 AM
post40889
|
Re: RE: mbuf leak (related to pppd)
But back to the original problem.
Now I have "home compiled" version, which leaks 1 mbuf at pppd start, and 2 mbufs when pppd terminates.
This is the information about the source code:
http://community.qnx.com/svn/repos/core_networking/trunk
ff24a42e-91d7-4d6b-b188-f2e9bd5c50fa
Rev. 2, author seanb
PKY
> Thanks for all the assistance.
>
> PKY
>
> > On Wed, Oct 28, 2009 at 02:53:07AM -0400, Pavol Kycina wrote:
> > > Thanks,
> > >
> > > Now I have compiled executable.
> > >
> > > But at the same time I have development system in "non defined" state, as
>
> > make install didn't work on stageing area, but on real system.
> > >
> > > I have followed:
> > > http://community.qnx.com/sf/wiki/do/viewPage/projects.core_os/wiki/
> > BuildTheOSSource
> > >
> > > this is the state of my system:
> > >
> > > # pwd
> > > /tmp/qnx_core_net2/lib/io-pkt
> > > #
> > > # set | grep QCONF
> > > QCONF_OVERRIDE=/tmp/qnx_core_net2/lib/io-pkt/qconf-override.mk
> > > #
> > > # cat qconf_override.mk
> >
> > You've exported 'qconf-override.mk' but setup 'qconf_override.mk'
> > ('-' vs '_'). This is why you shouldn't 'make install' as root
> > (at least the first time). :)
> >
> >
> > -seanb
>
>
|
|
|
Sean Boudreau(deleted)
|
Re: RE: mbuf leak (related to pppd)
|
Sean Boudreau(deleted)
10/28/2009 10:17 AM
post40891
|
Re: RE: mbuf leak (related to pppd)
On Wed, Oct 28, 2009 at 10:14:20AM -0400, Pavol Kycina wrote:
> But back to the original problem.
>
> Now I have "home compiled" version, which leaks 1 mbuf at pppd start, and 2 mbufs when pppd terminates.
>
You're sure you're actually running the one you
compiled? Using one at start is OK but it should
now be reaped on cleanup.
-seanb
|
|
|
Pavol Kycina
|
Re: RE: mbuf leak (related to pppd)
|
Pavol Kycina
10/28/2009 10:26 AM
post40894
|
Re: RE: mbuf leak (related to pppd)
I have triple checked, that I am running new compiled version.
> On Wed, Oct 28, 2009 at 10:14:20AM -0400, Pavol Kycina wrote:
> > But back to the original problem.
> >
> > Now I have "home compiled" version, which leaks 1 mbuf at pppd start, and 2
> mbufs when pppd terminates.
> >
>
> You're sure you're actually running the one you
> compiled? Using one at start is OK but it should
> now be reaped on cleanup.
>
> -seanb
|
|
|
Pavol Kycina
|
Re: RE: mbuf leak (related to pppd)
|
Pavol Kycina
10/30/2009 3:07 AM
post41033
|
Re: RE: mbuf leak (related to pppd)
Is your inhouse version working fine?
If so, could you please send me the binary?
Thanks, PKY
> I have triple checked, that I am running new compiled version.
>
> > On Wed, Oct 28, 2009 at 10:14:20AM -0400, Pavol Kycina wrote:
> > > But back to the original problem.
> > >
> > > Now I have "home compiled" version, which leaks 1 mbuf at pppd start, and
> 2
> > mbufs when pppd terminates.
> > >
> >
> > You're sure you're actually running the one you
> > compiled? Using one at start is OK but it should
> > now be reaped on cleanup.
> >
> > -seanb
>
>
|
|
|
Pavol Kycina
|
Re: RE: mbuf leak (related to pppd)
|
Pavol Kycina
11/02/2009 8:30 AM
post41144
|
Re: RE: mbuf leak (related to pppd)
Ping?
> Is your inhouse version working fine?
>
> If so, could you please send me the binary?
>
> Thanks, PKY
>
> > I have triple checked, that I am running new compiled version.
> >
> > > On Wed, Oct 28, 2009 at 10:14:20AM -0400, Pavol Kycina wrote:
> > > > But back to the original problem.
> > > >
> > > > Now I have "home compiled" version, which leaks 1 mbuf at pppd start,
> and
> > 2
> > > mbufs when pppd terminates.
> > > >
> > >
> > > You're sure you're actually running the one you
> > > compiled? Using one at start is OK but it should
> > > now be reaped on cleanup.
> > >
> > > -seanb
> >
> >
>
>
|
|
|
Sean Boudreau(deleted)
|
Re: RE: mbuf leak (related to pppd)
|
Sean Boudreau(deleted)
11/02/2009 8:56 AM
post41146
|
Re: RE: mbuf leak (related to pppd)
On Mon, Nov 02, 2009 at 08:30:44AM -0500, Pavol Kycina wrote:
> Ping?
>
> > Is your inhouse version working fine?
> >
> > If so, could you please send me the binary?
> >
Mine appears to be working fine. It should
be the same as what's on the foundry. How
exactly are you still creating a leak?
-seanb
|
|
|
Pavol Kycina
|
Re: RE: mbuf leak (related to pppd)
|
Pavol Kycina
11/02/2009 9:20 AM
post41150
|
Re: RE: mbuf leak (related to pppd)
I start io-pkt with these params:
io-pkt-v4 -dspeedo name=en -ptcpip -ppppmgr
And pppd:
pppd lcp-echo-failure 5 lcp-echo-interval 5 /dev/ser4 115200 nodetach defaultroute debug
There's nothing connected to ser4, so it timeouts.
I am monitoring mbuf count by netstart -m
When pppd starts, one mbuf is consumed and 2 mbufs are consumed "a second before" pppd terminates (probably doing some
cleanup before exiting).
As soon as pppd terminates, it's restarted with the same args.
It tekes approx. 1/2 minute for one "cycle".
Regards, PKY
> On Mon, Nov 02, 2009 at 08:30:44AM -0500, Pavol Kycina wrote:
> > Ping?
> >
> > > Is your inhouse version working fine?
> > >
> > > If so, could you please send me the binary?
> > >
>
> Mine appears to be working fine. It should
> be the same as what's on the foundry. How
> exactly are you still creating a leak?
>
> -seanb
|
|
|
|