Issue found in Armada 370: "No buffer space available" error during continuous ping

[Thomas Petazzoni]

Dear Maggie Mae Roxas,

On Mon, 7 Jul 2014 19:27:22 -0700, Maggie Mae Roxas wrote:
> Sorry, correcting typo:
> 
> - 3.13.9 (issue exists)
> - 3.10.24 (issue does NOT exist)
> - 3.13.5 (issue does NOT exist)

Ok, thanks again for the report. Unfortunately, you're hitting a known
problem that was fixed, but after the 3.13-stable cycle was closed.
Basically, between 3.13.5 and 3.13.9, I introduced two patches to the
network driver:

$ git slog --author=free-electrons v3.13.5..v3.13.9
396b229b683fdc08d8705883860ec5a1b810546a net: mvneta: fix usage as a module on RGMII configurations
ea64e1f33d9d627da5d38da035e5d7443276e84e net: mvneta: rename MVNETA_GMAC2_PSC_ENABLE to MVNETA_GMAC2_PCS_ENABLE

To fix the usage of mvneta on RGMII configurations. However, by doing
so, I broke SGMII configurations. So, I sent a patch to revert "net:
mvneta: fix usage as a module on RGMII configurations", which was
accepted. But in the mean time, the 3.13-stable cycle was closed, so
this revert was never merged in the 3.13.x series.

If you want to stay on 3.13.x, you should therefore apply:

cd71e246c16b30e3f396a85943d5f596202737ba Revert "net: mvneta: fix usage as a module on RGMII configurations"

This commit is from the 3.14-stable branch, in Linus master branch,
it's:

cc6ca3023f2c2bbcd062e9d4cf6afc2ba2821ada Revert "net: mvneta: fix usage as a module on RGMII configurations"

Some other commits, merged after that, fix the usage of mvneta as a
module on both RGMII and SGMII configurations.

Any reason you're still using 3.13.x ? You should really consider
switching at least to v3.14.x, which is a long-term version, and
therefore still maintained. If you use v3.14.x, all of your bug reports
that end up in patches, will ultimately be fixed in v3.14.x.

Thanks again for your report!

Thomas
-- 
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux, Kernel and Android engineering
http://free-electrons.com