boards: sam_e70_xplained: Enable L2 ethernet layer

[jukkar]

As we have E70 ethernet driver enabled, we need to enable the
generic ethernet L2 driver too so that the ethernet is actually
useful.

Signed-off-by: Jukka Rissanen [email protected]

[codecov-io]

Codecov Report

Merging #6789 into master will decrease coverage by <.01%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master    #6789      +/-   ##
==========================================
- Coverage   54.99%   54.99%   -0.01%     
==========================================
  Files         456      456              
  Lines       43842    43842              
  Branches     8225     8225              
==========================================
- Hits        24112    24111       -1     
- Misses      16386    16387       +1     
  Partials     3344     3344

Impacted Files	Coverage Δ
kernel/posix/pthread.c	69.5% <0%> (-0.5%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 87d6c50...22df3bc. Read the comment docs.

[jukkar]

This needs more thinking, lot of tests are failing here and there after activating L2 ethernet by default for E70.

[jukkar]

New version fixing sanity issues.

[jukkar]

Some of the sanity tests need bigger network buffer limits when compiling for E70 Xplained board.

[jukkar]

@nashif are you ok with this PR?

[nashif]

dont like this, just settle on a default that works for everyone, this is just a sample, so please try to have 1 prj.conf.

[jukkar]

I would have done that if that would have been easy. The E70 ethernet driver checks that there is enough network buffers available and gives error if not. Unfortunately we have several prj.conf files in samples/tests that have less buffers configured.
This means that we either

• Remove the checks in E70 eth driver. I would like to avoid this as these checks make sense for this driver

or

• We increase the buf counts in various samples. This is not good either as then some of the smaller memory boards will start to fail

or

• Blacklist E70 in some of the failing tests. This is certainly an option but I preferred to just increase the buf counts for these couple of tests so that we get to compile them for E70.

If you have some other idea, please suggest.

[nashif]

i would go with options 2:

We increase the buf counts in various samples. This is not good either as then some of the smaller memory boards will start to fail

[nashif]

ditto

[jukkar]

Increased the default number of buffers if ethernet is enabled. This allowed us to remove various net_buf configs from prj.conf files.

[jukkar]

@nashif does this look better now?

[pfalcon]

I missed this change. And it does not do what it advertises - it doesn't "Fix sanitycheck for sam_e70_xplained board". Instead, it works around bugs in the IP stack in a typical Zephyr way - by throwing large amounts of memory, to minimize probability of bugs and deadlocks of which the IP stack is full (#5857, etc.).

Neither it masks it only for sam_e70_xplained, it does it for all boards with NET_L2_ETHERNET, which affects quite a number of boards, starting with qemu_x86 (which uses SLIP, which is NET_L2_ETHERNET).

Some of the sanitycheck tests were having too small limit for
network buffers when compiling for sam_e70_xplained board.

Small limit of network buffers or large, it should work. Inability to work with small number of buffers points to bugs in the stack.

[pfalcon]

At the same time, some long-standing issues were actually resolved/alleviated. I was able to run net/sockets/big_http_download sample (transferring megabytes of data from the Internet) with quite minimal buffer settings with qemu_x86, and then was able to use the same settings with frdm_k64f and Ethernet: pfalcon#3 (comment) .

So, I'd suggest to revert this patch (specifically a0df4f6 commit), document what's wrong when running what samples on sam_e70_xplained, compare the behavior with other Ethernet boards like frdm_k64f, and then see if it's sam_e70_xplained specific issue, Ethernet specific issue, or bug in the IP stack.