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

[Thanks a lot to Enkhbold Ochirsuren of GSI who traced back a strange
behaviour to the "printf" incompatibility].

This removes "printf -v <outputvar> <format> ..." which is only
available in bash.

These needless incompatible extensions to standard commands are a
source of endless headaches. New features must spit an error when
unavailable (i.e., bash people should add vprintf, not break printf,
imho)

In this case, my previous "fix" to this file was incomplete, because
I fixed a bashism but didn't see this other one.

Also, this changes the shebang to call sh (which is already done by rcS:
only the interactive execution now uses sh instead of bash).

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

In sh, string comparison for equality is "=", not "==". Bash accepts
either operator but suggests to use "=" for posix conformance.
Our /bin/sh is currently busybox, which accepts "==" like bash, but
if you use ash or dash it fails:

   $ CONFIG_FOO=y
   $ if [ "$CONFIG_FOO" = "y" ]; then echo ok; fi
   ok
   $ if [ "$CONFIG_FOO" == "y" ]; then echo ok; fi
   dash: 3: [: y: unexpected operator

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

Output for the mask parameter was wrong if high bits are zero.
Not exposed by default (mask is 0xffffffff).

But I got

     VID    FID       MASK       DROP    PRIO    PRIO_OVERRIDE
    102    102      0x   7ffff    NO      --         NO

Which now is

    102    102      0x0007ffff    NO      --         NO

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

This adds the git headers, so we can easily rebuild a buildroot-for-wrs
tree to add more patches.

This is how I do it now (where "wrs-build" and "wr-switch-sw" are
at the same directory level.

   cd wrs-build

    mkdir tmp; cd tmp
   tar xf ../downloads/buildroot-2016.02.tar.bz2
   cd buildroot-2016.02/

   git init; git add .; git commit -m "buildroot-2016.02 as untarred"
   git am ../../../wr-switch-sw/patches/buildroot/00*
   mv .git ../../build/buildroot-2016.02/

   cd ../../build/buildroot-2016.02
   <work-here>

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

There is a monit-related problem in commit 9bc9dd16

.

The commit renames /etc/init.d/lldpd.sh into /etc/init.d/lldpd.bash
whle adding a single stanza. Try "git diff e61eb08 9ebc5e2" to
see the difference.

But monit still uses the old name. Instead of changing monit
(a two-liner, but adds noise because of .bash), this renames back
the file to .sh and removes the needless bashism in the new stanza.

Please note that calling stuff ".bash" makes no sense (suffixes
make no sense most of the time!). The only thing rcS does is checking
if the name is "*.sh", to just read it in the current shell instead
of executing. So anything not ".sh" is executed, we don't need to
special case .bash, .py or whatever.

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

RX LLDP frames worked without any change.

Signed-off-by: Michal Wasiak <michal.wasiak@gmail.com>

Signed-off-by: Kinga Wujek <dev_public@wujek.eu>

Signed-off-by: Kinga Wujek <dev_public@wujek.eu>

Note:
--Checking of checksums was added in v5.0 of firmware
--Added note about limiting the size od LLDP frame

Signed-off-by: Kinga Wujek <dev_public@wujek.eu>

Signed-off-by: Kinga Wujek <dev_public@wujek.eu>

see: ppsi#25

Maciej Lipinski authored 4 years ago and Grzegorz Daniluk committed 4 years ago

softPLL was set in BC mode and current_reference changed during
the measurment. The problemw as that the before-measurement
setting was not recovered. This caused problems, in particular
during LPDC rx calibration of other ports.

Maciej Lipinski authored 4 years ago and Grzegorz Daniluk committed 4 years ago

printf makes the messages to be logged in syslog in strange order
with respect to the other messages. It's best to use pr_* consistently.

Nico Coesel authored 6 years ago and Grzegorz Daniluk committed 4 years ago

NC: Add patches for 1GB NAND flash on SCB3.4.1. These patches are (tested) backwards compatible with the original SCB3.4.

Maciej Lipinski authored 4 years ago and Grzegorz Daniluk committed 4 years ago

In the rx calib FSM, this is needed to over the case when the
calibration is re-started (INIT state entered) during execution
of the FSM (e.g. due to link unplugging, or timeout).

In the tx calib FSM, in theory this is not needed, yet it seems
clean and it might be needed in the future if some sort of
reset is implemented.