Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

The USB gadget framework has changed from Linux v2.6.39. As far as I
understood it's now more modular, so it's not anymore enough to load
just the g_serial.ko driver. Now we have to load also it's
dependencies.

I'm not an expert about USB gadget, this patch is the result of my
understanding in debugging the loading failure of the g_serial.ko
module. Perhaps it will be necessary some adjustments to make it
work the way we want. But for the time begin I will leave it like
this until is clear what we have to adjust.

Note that these dependecies depends also on the configfs file-system.
I had problems to get this work because apparently by using a different
corss-compiler some symbols where removed from the kernel, so our
modules were not able to load because of "undefined symbols".
Using the toolchain compiled within this project solved this issue.
Signed-off-by: Federico Vaga <federico.vaga@cern.ch>

The kernel version is hardcoded here. Take it at runtime by using
`uname -r`. This will reduce incompatibilities on kernel updates
Signed-off-by: Federico Vaga <federico.vaga@cern.ch>

Check md5sums of files in the wrs-firmware.tar with stored in file
checksums.md5. When checksum does not match abort the update.
Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

When switch's firmware with buildroot older than 2016.02 was updated via
copying wrs-firmware.tar with buildroot at least 2016.02 to the
/update partition, then hwinfo in the flash was overwriten.
Such behaviour was caused by crashing during execution of /wr/bin/sdb-read.
Crashing sdb-read caused the following if's condition from the file
/etc/init.d/hwinfo to be false.

if /wr/bin/sdb-read /dev/mtd5ro > /dev/null; then

This triggered later overwrite of hwinfo partition.

/wr/bin/sdb-read was crashing because during update the binary /wr/bin/sdb-read
was called from just extracted firmware. Problem pops up when libraries
required by sdb-read are in the different versions in the old and the new
firmware. Unfortunatelly, this problem cannot be easily solved, since we cannot
change the update procedure in already installed firmware (the one that is
performing update).

Implemented solution does the following:
--change the name of /wr/etc/sdb-for-dataflash to /wr/etc/sdb-for-dataflash.bin
  by this /etc/init.d/hwinfo script exits with an error without modification
  to the hwinfo partition.
--rename link /etc/rcS/S90hwinfo to /etc/rcS/S10hwinfo, which will trigger
  update of hwinfo partition earlier during the boot.
--restart switch after update of hwinfo in the flash
--remove calls of /etc/init.d/hwinfo during the update
Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

The problem is that the "reboot" command sends the signal to the init. However,
it looks like init queues all incoming signals. They're handled after all
"sysinit" commands from the inittab are finished.
reboot -f skip sending the signal and triggers the reboot immediately.

This problem pop up after update of a buildroot to version 2016.02.
Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

Move:
/etc/init.d/[SK]??<name> to /etc/init.d/<name>
Then create symlinks in
/etc/init.d/rcS/[SK]??* to all files that previously were /etc/init.d/[SK]??*
With exception of monit since monit will start soon from init.

Rename all occurences of changed files.
Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

Note: Be aware that date of last update may be not accurate.
One reason is that it uses date stored before last restart, so time needed to restart is not taken into account.
Additionally date stored before restart may be not accurate for many reasons like:
--hard restart (last know date was not saved)
--date was never received from ntp nor set manually
Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

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

If the usb cable is disconnected, any write to /dev/ttyGS0 goes to a
buffer until it fills. Then, any further write will block. This means
the messages towards the user ("uncompressing filesystem... done" etc)
caused the update procedure to stop.

This fixes the problem by using non-blocking write for the messages:
if nobody`s reading the messages are just discarded by the kernel.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

I was bored by repeated file names, so this uses WRS_KER and other
variables to host the "magic" name.

Similarly, the use of "-old" was inconsistent, and this uses the
same pattern for all files. Actually, it should be a function
to avoid repetitions. But I'm lazy and it is not.

The upgrade we may do now is by placing files in /update, with
magic names:

    - wrs-firmware.tar: is *must* include wrs-usr.tar.gz, because
      if this file exists /usr is recreated from scratch.

         - it may include zImage, wrs-initramfs.gz, barebox.bin
	 (this happens in released firmware files, so the boot loader
	 is being written twice, until we get rid of the sam-ba tools)

    - zImage: if present, the kernel is replaced

    - wrs-initramfs.gz: if present, the initramfs is replaced

    - barebox.bin: if presente, the boot loader is replaced

This means, in practice, that you may change the whole system or one
part at a time.  I suppose only we developers will use the "one part
at a time".

But the most important change is that barebox.bin is being replaced
during boot.  In v4.0 we didn't expect to ever replace barebox, but
then we found we needed it -- so let's provide for further updates.
OTOH at91boot is not replaced, and it is not expected to, ever.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

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

This reverts commit 38e59051.

We don't need the sdram printouts any more. But it's good to
have this in history anyways (it was an error having it in master,
but all hackers promote bugs as features).

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

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

We had
   /etc/init.d/wrs-boot-procedure: line 104: tee: not found

because tee is not in /bin; we need to always use "busybox tee".
This adds a variable TEE="busybox tee" to make it easier to remind, at
later edits, and fit in 80 columns at the same time.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

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

Thanks to Cesar Prados for finding 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>

This brings everything into /etc/init.d/wrs-boot-procedure that is
executed as a "sysinit" task, before anything else.  The procedure
runs from the initrd/initramfs and is concerned with the complete
update of the device, if an update is pending.

See documentation (next commit) for details about partitions and boot
procedure.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>