Hello Niko,

we solved the memory leak for our project like so:

Problem:

From: Michael Reeder <m...@datatroniq.com>
To: etherlab-users@etherlab.org
Subject: [etherlab-users] unofficial ec_igb driver for kernel 4.19 from Gavin Lambert likely leaks memory
Date: Tue, 14 Jan 2020 12:01:15 +0100

Hi All,

we are using the Etherlab master with the newest patches from Gavin Lambert (changeset 12:8ea39404a2e8, http://hg.code.sf.net/u/uecasm/etherlab-patches/rev/8ea39404a2e8 ) which include a native ec_igb driver for kernel 4.19.

Unfortunately I found that when starting the Etherlab master with this driver, free memory on the system is constantly decreasing. This doesn't happen when the generic driver is used we assume the igb driver has some memory leak.

Did anybody also uses this driver and experienced the same behavior?


Solution:

From: Michael Reeder <m...@datatroniq.com>
To: Gavin Lambert <gavin.lamb...@tomra.com>, Gary Grobe <ggr...@houstonmechatronics.com>, etherlab-users@etherlab.org
Subject: Re: [etherlab-users] ec_igb memory leak?
Date: Fri, 13 Mar 2020 09:37:55 +0100

Hi,

I had the same issue with the 4.19 ec_igb driver.
When comparing the sources with 4.4 it turned out that a line is missing in `devices/igb/igb_main-4.19-ethercat.c`

8475| /* retrieve a buffer from the ring */
8476| if(adapter->ecdev) {
8477| unsignedchar*va = page_address(rx_buffer->page) + rx_buffer->page_offset;
8478| unsignedintsize = le16_to_cpu(rx_desc->wb.upper.length);
8479| ecdev_receive(adapter->ecdev, va, size);
8480| adapter->ec_watchdog_jiffies= jiffies;
*+++ **igb_reuse_rx_page**(rx_ring, rx_buffer); *// copied this line from igb_main-4.4-ethercat.c -> fix memory leak

8481| }

Adding this line from 4.4 source fixed the leaking for me. (line is also missing on 4.14)

Please note that I didn't really study the code besides comparing it to 4.4. I do not know if this is safe to do!


Hope this may help
Michael


Best, Jürgen

On 22 Feb 2021, at 8:01, Kivel Niko (PSI) wrote:

Hi Laurent


Did you by any chance also make the 'igb' driver work with 5.4? I noticed a memory leak in 3.10 and also in 4.19. I recall there was discussion about this year ago, but lost track of it.


best

Niko


--
Paul Scherrer Institute
Dr. Niko Kivel
Forschungsstrasse 111
WBGB/017
5232 Villigen PSI
Switzerland

phone: +41 56 310 2226
email: niko.ki...@psi.ch
________________________________
From: Etherlab-users <etherlab-users-boun...@etherlab.org> on behalf of Karl Zeilhofer <k...@zeilhofer.co.at>
Sent: Sunday, February 21, 2021 12:02:18 AM
To: etherlab-users@etherlab.org
Subject: Re: [Etherlab-users] [etherlab-users] Compatibility with kernel 5.x

Dear Laurent,

did you do any further upgrades?

Do you run the master in kernel or user mode?

What exactly was the cause of the problem?

Kind regards,
Karl

On 6/19/20 2:44 PM, Joye Laurent wrote:
I found and fixed my problem!
It was a kernel configuration failure.

My system is also working fine with a kernel 5.4.

So sorry for this disturbance !!

Regards
Laurent

On Fri, 2020-06-19 at 08:57 +0000, Joye Laurent wrote:
Hello,

I'm using the Etherlab master with the patches from Gavin Lambert
(changeset 33b922ec1871
http://hg.code.sf.net/p/etherlabmaster/code
)
with a kernel 4.19. It's work like a charm!

I have to upgrade my system
to fix the Y2038 problem. That's why I
need to upgrade to the Linux kernel 5.4.

But Since I boot my system with the kernel 5.4, the EtherCAT master
doesn't found anymore slaves. There are no announced error.

Did anybody already made some experiences with a Linux kernel > 5?

Regards
Laurent
_______________________________________________
etherlab-users mailing list
etherlab-users@etherlab.org

http://lists.etherlab.org/mailman/listinfo/etherlab-users





_______________________________________________
etherlab-users mailing list
etherlab-users@etherlab.org
http://lists.etherlab.org/mailman/listinfo/etherlab-users


--
Etherlab-users mailing list
Etherlab-users@etherlab.org
https://lists.etherlab.org/mailman/listinfo/etherlab-users
--
Etherlab-users mailing list
Etherlab-users@etherlab.org
https://lists.etherlab.org/mailman/listinfo/etherlab-users

Reply via email to