This is the mail archive of the ecos-discuss@sources.redhat.com mailing list for the eCos project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Problem with mbufs...


Hi Prem,

It may be possible that this warning is not really a code problem but more of a performance issue. If you stop flooding the target with pings and send them normally does the problem still occur?

Later,

Steve

M.PREM KUMAR wrote:

Hi,

I updated to latest version of ecos2.0 (16-12-02) taken source from
ecoscentric and ported on our custom board (MPC860 based). Still we are facing the problems with mbufs especially when we give
flood ping of size 1500 (pkt size).
eg. ping -f -s 1500 <ip-addr> will immediately start giving
"warning: eth_recv out of mbufs".


regards
prem.




On 18 Dec 2002, Gary Thomas wrote:


On Wed, 2002-12-18 at 09:29, M.PREM KUMAR wrote:

Hi Andrew,

I am using ecos version 1.3.1

^^^^^^^^^^^^^^^^^^

Here's your problem! This code is terribly old and the network
stack has lots of problems, especially in the area of memory management.

Update to the latest CVS and your problems will probably go away.


I made CYG_PKG_MEM_USAGE to double the size of default value. Still
there was no use.

regards
prem.





On Wed, 18 Dec 2002, Andrew Lunn wrote:


On Wed, Dec 18, 2002 at 06:04:50PM +0500, M.PREM KUMAR wrote:

Hi Andrew,

First let me thankyou for the information provided.
I am not using latest version of eCos. Is there any bug w.r.t mbufs in
older version?

Depends on how old, but yes.


Further, I am sorry I couldn't get your question, "Checking of
receiver half of the code is not leaking as well"
Can you kindly elaborate it.

ping sends and receives. You say you have checked eth_drv_tx_done. So
you have checked for leaks in mbufs used for transmitting. Have you
looked for leaks of mbufs when receiving the ICMP ECHO reply?


Current value of CYGPKG_NET_MEM_USAGE is 262144 (Default value).
This controls the number of mbufs, clusters and memory pool. See
packages/net/tcpip/current/src/ecos/support.c If you want more mbufs,
increase this.

Also, please CC ecos-discuss@sources.redhat.com

Andrew


rgds
prem





On Wed, 18 Dec 2002, Andrew Lunn wrote:


Are you using the latest version of eCos from anoncvs?


I am running ping program on our card (Uses MPC860 as base
processor). When trying to send more pings, there is a warning of
"eth_recv out of Mbufs" and I am not able to send further pings. Ethernet
driver is calling eth_drv_tx_done() properly for freeing resources (in
this case
mbufs.)
Have you checked that the receiver half of the code is not leaking as
well?


I want to give a try by increasing amount of Mbufs. Can anyone tell how
to increase the amount of MBufs. Further, if anyone already had a similar
problem, provide with the necessary information to solve.
Increase CYGPKG_NET_MEM_USAGE.

Andrew



--
Before posting, please read the FAQ: http://sources.redhat.com/fom/ecos
and search the list archive: http://sources.redhat.com/ml/ecos-discuss



--
Before posting, please read the FAQ: http://sources.redhat.com/fom/ecos
and search the list archive: http://sources.redhat.com/ml/ecos-discuss

--
------------------------------------------------------------
Gary Thomas |
MLB Associates | Consulting for the
+1 (970) 229-1963 | Embedded world
http://www.mlbassoc.com/ |
email: <gary@mlbassoc.com> |
gpg: http://www.chez-thomas.org/gary/gpg_key.asc
------------------------------------------------------------







--
Before posting, please read the FAQ: http://sources.redhat.com/fom/ecos
and search the list archive: http://sources.redhat.com/ml/ecos-discuss


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]