Before the cleaning was not done and the git tag was not taken into
account if modified without modifying any source.

The package will be named as following: <tag_name>-<date>_binaries.tar.gz
If the <tag_name> does not start with the "official" wr-switch-sw- prefix
a forced prefix 'wr-switch-sw-unof-' will be added.
Examples:
  * wr-switch-v3.3.1 => wr-switch-sw-v3.3.1-20130902_binaries.tar.gz
  * v3.1-rc1         => wr-switch-sw-unof-v3.1-rc1-20130521_binaries.tar.gz

Also replace the generated .bin in the binaries/ folder

undocumented hack to enable v3.3 lx240t scb on v3.2 miniBP only by
creating an empty file named /wr/etc/lx240t.conf. This is only
needed for v3.2 miniBP because they do not have the i2c chip and its
DIP switch to setup version.

When we had a pulldown on the i2c bus the device was okay because we
only check ack (0), now we first check if we have a pullup and then a
ACK.

This is a patch of patch applied to m4 library during buildroot compilation.
This error occurs with recent GCC version (4.7.3 on ubuntu 13.04).

It solve the error: "can't create directory '/dev/shm': Read-only file system"
when /etc/inittab want to create /dev/shm folder. 
This happens since FS is read-only.

Git revision can refer to the at91bootstrap repo itself used to generate
the patches (in this case .git exist) or to a parent repo (wr-switch-sw)
that we used with patches (In this case we need to pass git value using
exported variable) because no .git folder when creating the
at91bootstrap folder.

WR_INSTALL_ROOT and WR_LIB are folders to install library while using NFS.
They should only be used with install command and not during linking step
because they can be out of sync.

The HDL sources can be found in wr-switch-hdl repo with the tag wr-switch-htl-v3.3

On some swtiches the management port was not accessible, most likely
for wrong timing in overall setup.  Removing and re-loading the
usb gadget module seems to fix the problem.

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

v3.3 gateware has reduced number of global clocks by using regional clock nets
for first 4 GTX. Software needs to compensate additional delays.

In Grand Master mode it takes longer to lock PLLs to external source, therefore
HAL becomes ready later.

The script wrfs_mnt can be used for remounting rootfs to update the
configuration of the switch.

0. From power until linux has boot the led is unpowered
1. We first set init it before flashing FPGA => LED is yellow.
2. Then when the HAL start => LED is orange
3. WR is setup => LED is green

Despite being "obivously right", the previous commit, didn't work.
I didn't identify the reason of the permission problem but changing
it right before tarring works.

Previous commit is already in master, and master is holy, so here's
another commit.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Don't filter by VID mask on. Such filtering is not a default 802.1Q behavior. TODO: add optional config.

Override dst_port_mask when ureq received and there is existing MAC Entry. Update time when updating existing entry