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>

The interrupt discussion has been moved to the projects' wiki.
Besides that, some typos, external references and nothing else.
And today is not yesterday: new date is there.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This is mainly due to feedback by Matthieu Cattin and Wesley Terpstra
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This adds a section that explains why we don't describe interrupts.
Most of the section comes from a message by Wesley Terpstra,

  http://lists.ohwr.org/sympa/arc/fpga-config-space/2013-03/msg00000.html

(you may need to re-enter the same url in the browser after clicking on
the "I'm not a spammer" button).

I'm not personally able to make the discussion shorter, because I think
it's very informative.  The message has been slightly reworded to make
it less informal and more suitable to be an "official" rationale, and
I added some glue lines at the beginning and the end.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

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

With the new forced page breaks (previous commit) two pages had just a
single line in them. So this shrinks a little the text to fix the
problem.  I don't want to force page breaks only in some of the
subsections, for consistency.

Note: I tried the argument "3" to page breaks, to have them strongly
"suggeested" but not "mandated" (argument 4 or default if no argument),
but the result was like without page breaks (i.e., bad).
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This is useful so the image and the table of fields are always on the
same page.  The default effect with float figures and table was
difficult to use as a reference while writing code.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This adds the "repo_url" and "synthesis" records, that are already
used in one of your projects.

Thanks to Matthieu Cattin for his help in designing the structures
and actually implementing them.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

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

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

By moving the integration description after device and bridge,
we describe the records in the numerical order, which is also
logical because the integration is only an informative record.
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>

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 is a merge of all previous work done on sdbfs using a different
repository (which I used to push to gitorious).

Besides the kernel stuff, which is still in flush, userspace and lib
are actually useful to sdb users in general, so this is a better place
to host it.

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

According to the kernel headers you have, compiling the kernel part
may fail. So compile userspace first, in case you need gensdbfs alone.
Thanks to Matthieu Cattin for noting this.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

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

This a major thing, that changes a lot. However, now directories work
properly and the code is generally cleaner. This is what I get,
for example, for the wr-nic release gateware on a SPEC:

  spusa.root# find /mnt
  /mnt
  /mnt/WB4-Crossbar-GSI
  /mnt/WR-CORE
  /mnt/WR-CORE/WB4-Crossbar-GSI
  /mnt/WR-CORE/WB4-BlockRAM
  /mnt/WR-CORE/WB4-Bridge-GSI
  /mnt/WR-CORE/WB4-Bridge-GSI/WB4-Crossbar-GSI
  /mnt/WR-CORE/WB4-Bridge-GSI/WR-Mini-NIC
  /mnt/WR-CORE/WB4-Bridge-GSI/WR-Endpoint
  /mnt/WR-CORE/WB4-Bridge-GSI/WR-Soft-PLL
  /mnt/WR-CORE/WB4-Bridge-GSI/WR-PPS-Generator
  /mnt/WR-CORE/WB4-Bridge-GSI/WR-Periph-Syscon
  /mnt/WR-CORE/WB4-Bridge-GSI/WR-Periph-UART
  /mnt/WR-CORE/WB4-Bridge-GSI/WR-Periph-1Wire
  /mnt/WR-CORE/WB4-Bridge-GSI/WB4-Streaming-DMA_0
  /mnt/WR-NIC
  /mnt/WB-VIC-Int.Control
  /mnt/WR-TXTSU
  /mnt/WR-DIO-Core
  /mnt/WR-DIO-Core/WB4-Crossbar-GSI
  /mnt/WR-DIO-Core/WR-1Wire-master
  /mnt/WR-DIO-Core/WB-I2C-Master
  /mnt/WR-DIO-Core/WB-GPIO-Port
  /mnt/WR-DIO-Core/WR-DIO-Registers
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>

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>

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

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

If the host is little-endian and the bus is directly connected as 32
bits, byte access would be wrong.  This commit detects the thing from
the magic number value and reversed the words of the SDB records.
Missing that, strings would be reversed and 16-bit fields would be in
the wrong byte offset.  Please note that the resulting SDB record
still has big-endian fields.

This problem results from bit 0 being mapped to bit 0 and bit 31 being
mapped to bit 31 (i.e., 32-bit words are native mapped on both sides
of the bridge).  However, on the sdb bus bit 31 (MSB) lives at byte 0,
while the little-endian host stores it at byte 3.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>