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

This reverts commit 03062ea3.

The patch is wrong, because it duplicates

    9491e61b

 ppsi: pick new master and fix building accordingly

which does the "fixing" in a more elegant way. With both in history,
we now have

   ./ppsi/include/generated/autoconf.h:13:0: warning:
           "CONFIG_VLAN_ARRAY_SIZE" redefined
   generated/autoconf.h:14:0: note: this is the previous definition

Besides, the reverted commit was not even using new ppsi master (we
did it before this commit).

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

Commit 70b42d0d is currently proposed_master, so by our agreements with
GSI we can't remove it from history.  The change itself is already
in master (but with typos fixed in the commit message). So let's
pin the commit by merging, even if it will appear twice in the history.

This just adds "void" in prototypes. I did the pfilter on a different
branch, and lost these changes of commit be2bd234

 wyhile moving pfilter
generation to tools/

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

Use the rules from the array we are passed in the object file.
We'll use the feature when enabling vlans: at that point we'll have
two sets to choose from, according to run-time configuration.

Using the build-time configuration avoids the hairy ifdef and saves
almost 1kB of binary size.

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

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

This changes pfilter-builder.c so it really build in user space, but removing
register setting and creating output files instead.

Running this program creates 3 files, that correspond to the previous
three #ifdef conditions.  The files are:

      rules-plain.bin    this is used for the no-etherbone configuration
      rules-ebone.bin    etherbone setup
      rules-e+nic.bin    etherbon plus 7solution's wr-nic packet filter

The files begin with a magix 0x11223344 word, that allows the soft-core
to fix any endianness difference (so no hairy mishaps are expected when
switching to a different soft-core). Then the 40-bit command words are
saved as 64-bit vaules, LSB-first. The output file is thus an odd number
of words and no 64-bit alignment is required.

The first three instructions of the packet filter are used to compare
the destination mac address of the frame.  We now use a fake mac address,
and the LM32 code will change it while programming the binary.

Please look at this commit while ignoring white-space, as it changes
indentation while turning #ifdef/#else into if()/else .

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

We are going to have a user-space tool to build the packet filter
images. The first step is copying dev/ep_pfilter.c, so we can
"git diff" the changes.  At this point the file is not
even built.

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

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

Simplify access to $CONFIG_RAMSIZE. Actually, the previous way
was not working for me, as my $SHELL is not bash.

The bug is uncovered by the next commit, because the "-s" argument
to genramvhd has always been wrong for me, but I didn't notice
(so maybe the "-s" argument has no effect? I can't tell at this point)

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

If we compare for the local MAC first, it's easier to patch
the mac address at run time.

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

We now explain how the class bits are used after the pfilter is over.

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Acked-by: Grzegorz Daniluk <grzegorz.daniluk@cern.ch>

There is no need to assign register names to numbers explicitly,
because there's always the risk to make errors.

This also adds a build-time check for any overflow of the 22 registers:
if we allocate more than 22, it won't build.

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

This removes the register numbers from comments, and rewords
comments in general to match the fact that registers are now named.

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

The current code was reusing R10 and R11 for a different role.
However, we have several free registers, so this allocates two new
more, by giving them a meaningful name.

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

Uses of R20 and R12 are temporaries, and very local. Thus,
we can reuse R14, already blessed R_TMP.

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

The existing code used R6 both for ARP and for ETHERBONE.  Allocate a
new register, since we have plenty. This clarifies the code (by using
proper symbolic name) but changes the generated binaries.

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

This commit is not changing the generated code at all (the
bits generated are exactly the same as in the previous commit),
in all three cases of #ifdef.

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

By naming registers the code is much more readable in my opinion.
And we avoid, in future modifications, to mistake a frame offset for
a register.

The enumeration of names starts from 1024 to deny use of simple
numbers. Also, the enumeration is verified to be exactly 32 names,
by raising a build error if not.

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

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

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

"act accordingly" means pre-including ppsi/...autoconf.h within
CFLAGS, or we miss CONFIG_MAX_VLANS_PER_PORT, used in defining the
ppsi local data structure.

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

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

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

The current default is a multicast address. We'd better use a
locally-assigned number instead. We keep the pattern of a very-visible
thing, so if it appears in some sniffer or ARP table you notice.

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

Because the golden bitstream for SPEC has syscon at offset 0x0.

E.g. Kintex-7

Based on Peter's work for km2net CLB board.

Reason: h_y was close to the DAC range, on some boards we were unable to lock
HPLL.
In addition to HPLL modifications, also the phase shifting had to be changed.
When offset clock has lower frequency than the ref clock, shifting ref
clock produces sampled clock which is shifted in another direction.