[Bug 1001344] Allow nc_test_slave for lwIP to compile for targets with 128KiB RAM.

[bugzilla-daemon@bugs.ecos.sourceware.org]

Please do not reply to this email. Use the web interface provided at:
http://bugs.ecos.sourceware.org/show_bug.cgi?id=1001344

--- Comment #18 from Ilija Kocho <ilijak@siva.com.mk> 2012-05-15 10:55:02 BST ---
(In reply to comment #16)
> (In reply to comment #15)
> > (In reply to comment #14)
> > > Here I am again after long time. It seem that buffers were too short
> > > 8KiB and messages can be up to 10KiB. Now I fixed this and it is
> > > stable. Unfortunately I can't test on synthetic target since newer
> > > Linux kernel (3) doesn't seem to support ethertap (any suggestion?).
> > > Sergei do u have some time to test on synthetic target pse?
> > >
> > > Note: The same bug with MAX_BUF seem to be present in BSD test too.
> > > I'm going to open a bug for that.
> 
> [snip]
> 
> > I will test your newest patch on synthetic target for 2.6.32 kernel
> 
> Ilija, I have tested nc_test_{master,slave} for synthetic target and NET
> template (all NC tests passed as I could expect). The eCos settings were
> 
>   cdl_configuration eCos {
>       template    net ;
>   };
>   cdl_option CYGPKG_NET_BUILD_HW_TESTS {
>       user_value 1
>   };
> 

Interesting. On my target /TWR-K70F120M NET/ test crashes (ref Bug 1001586)
unless patched http://bugs.ecos.sourceware.org/attachment.cgi?id=1755.
It is interesting that it doesn't crash on the initial test, that actually uses
10KiB messages, but later.

Just to note that the buffer enlargement is part of Attachment 1754 as well.

> Hm, not patched lwip' NC test (slave) on synthetic target stalled when
> background load is 30% (for BSD stack all loads passed on eCos synthetic
> target)
> 

Perhaps patch similar to http://bugs.ecos.sourceware.org/attachment.cgi?id=1755 
(Bug 1001586) could help for not patched lwIP.

As for Attachment 1754 it reduces thread stacks and I suspect that synthetic
target has larger context than Cortex-M.

Also by experimenting I have found out that some user settings are needed on
lwIP in order to work-out the test well (at least on Cortex-M). Pse see ECM
snippet below, might help with synth. One notable thing is to set Ethernet
thread the same priority as TCP/IP thread. I guess this setting reduces the
thread switches rate - by preventing Ethernet thread to interrupt TCP/IP
thread.
Also CYGNUM_LWIP_TCP_MSS is adapted to test conditions in order to avoid
message fragmentation.

------ lwip.ecm ----------
cdl_configuration eCos {
    description "" ;
    template    default ;
    package CYGPKG_IO_ETH_DRIVERS current ;
    package CYGPKG_NET_LWIP current ;
};

cdl_option CYGFUN_LWIP_MODE_SIMPLE {
    user_value 0
};

cdl_option CYGFUN_LWIP_MODE_SEQUENTIAL {
    user_value 1
};

cdl_option CYGNUM_LWIP_TCP_MSS {
    user_value 1200
};

cdl_option CYGDAT_LWIP_ETH0_DEFAULT {
    user_value 1
};

cdl_option CYGNUM_LWIP_MEM_SIZE {
    user_value 3200
};

cdl_option CYGNUM_LWIP_THREAD_ETH_PRIORITY {
    user_value 8
};

----- End of lwip.ecm ----------------

-- 
Configure bugmail: http://bugs.ecos.sourceware.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.