Grzegorz Daniluk authored 11 years ago and Alessandro Rubini committed 11 years ago

As it was done also in ptp-noposix. We need to try multiple times
because sometimes it takes time for HAL to become ready (especially in
the GrandMaster mode when it has to lock SoftPLL to an ext clock).

Pietro Fezzardi authored 11 years ago and Alessandro Rubini committed 11 years ago

given that INTS() is defined as inline it should not affect
the performance even in time-critical cases.

previously one could use config options without arguments, or
config options with wrong names, and the program did not complain.
now you have to use correct names for config options and only
options with type ARG_NONE can be without arguments.
hardcoded default config file and config string are supposed
always correct so there's still no error check on their parsing.
however for any config options passed from command line with -f or
-C all the checks are enabled. if there are some syntax errors or
options without value the process exits with errors, printing
informations on all the found configuration errors.

both in comments and in diagnostics

instead of bare numbers

now it reflects the definition in pp_ext_hooks. With this new prototype
you can always choose a different pp_runtime_opts to be used for the
initialization

instead to rewrite every time the checks for the parser

Every case of the switch was colling the cfg_handler out of the
switch construct, avery time with the same arguments.
The only case not calling the config function was case ARG_NONE.
But this doesn't make sense, because if a configuration option
does not do anything, then it should not be there, even without
arguments.
So the call to config function is now made from every case, even
ARG_NONE, then it should not stay inside the switch.

and all the data structures needed to set configuration options
about time.
In ppsi.h added struct pp_cfg_time a platform independent timespec-like
data structure.
union pp_cfg_arg is extended with a pp_cfg_time field.
enum pp_argtype is extended with ARG_TIME.
In lib/conf.c a case ARG_TIME is added to handle it.

in ppsi.h. This allows to pass only a single argument to cfg_handlers
configuration functions.
Definitions of functions and arguments have been updated in lib/conf.c to
follow this changes.

- backup the existing .config file, if present, and restore it before exiting
- make clean before exiting

Since check_packet is using network channels (file descriptors), it is
better declared as a network operations.  This allows, for example,
to make a simulator using ARCH=unix and only changing TIME= .

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

This simplifies the build system, by relying in weak symbols instead
of the link order. We need weak symbols anyways for configuration, so
we can get rid of the libraries.  The size of binaries returned at
build time is increased, because arch/ and lib/ stuff is not part of
ppsi.o (as it actually should).

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

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

The tool now builds all configs/* setups, but it has fewer options than
it used to have (no command-line arguments as yet, so "./MAKEALL bare"
will not just check bare configs.

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

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

This adds the needed configuration items: architectures, extension,
cross_compile and wrpc_root.  So, what we passed on environment and
command line is now in the configuration file.

This changes the priorities of such options, so the Makefile is
changed to force $(CC), $(LD) and $(CROSS_COMPILE) down to subprojects
(i.e. pp_printf and arch-wrs/mini-rpc).  Also, we need to remove
quotes from .config items, and this is ugly linesin ./Makefile

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

Now Kconfig works, but we have no active configuration at all.

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

Add the lines from the .gitignore found in bst-kconfig. As in the
previous commit, use bst-kconfig commit v2.6.34-4-g8168c0b

(cherry-picked from wrpc-sw d923523a)

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

This is the combination of the following commits (oldest on top)
from wrpc-sw:

  89cdfcd added kconfig infrastructure, the trivial way
  5fda0c8 kconfig: use pathnames from Linux-2.6.34
  ab30804 kconfig: use 'Kconfig' as file name, not the bst-kconfig ones
  7c78880 kconfig: trivially write "WRPC" in place of "Linux Kernel"
  2c30cb4 kconfig: trivially fix most 'kernel' words in messages
  70eb3ba kconfig: add a configs/ directory and use it
  64cc10f scripts/kconfig: don't complain for default in choices

After picking them, I replaced all "WRPC" with "PPSi", and I removed
the configs/  files (unrelated to this project).

In practice, at this point we have the Kconfig infrastructure of
Linux-2.6.34, which I picked from the "bst-kconfig" package, which did
the extraction work.

One day I forward-ported this from v2.6.34 to v3.4, by applying a few
hundred kernel patches. While the final code was working, we decided
the port was not needed for wrpc-sw. Here in PPSi I keep the same version
we have in wrpc-sw.

As of this commit, the code is not used; the default configuration
file and Makefile targets are added in the next commits.

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

Bug introduced by

   ce8485bf

 bmc: remove field ppi->frgn_rec_i

I forgot to run ./MAKEALL during the last commits.

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

This is needed for the wrpc-sw release, as Etherbone won't fit with
a full ppsi in the default RAM size for SPEC cards.

ptpdump support will soon become optional in ppsi, through Kconfig,
and a bigger RAM size will become the default for SPEC images as well.
But currently we need this hack.

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 was already expexted to happen, but it was done in the wrong
place (my fault).  Most likely we should move this
"enable_timing_output" in the ptp-wide operations, to be used
outside of WR as well.

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

The field is a duplicate of ppi->frgn_rec_num, and got out of sync.
This removal fixes WR syncronization as a slave.

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

Also, report how may are there in bmc diagnostics

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

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

This works around a design problem that must be checked carefully.
The "is from current parent" flags set at receive time reveales bad,
as we should only check when needed.

So by now just say it is the current parent if we don't have a current
parent yet (i.e., we are not in error, even if it is not from current
parent).

Also, forget about current parent and foreign masters when the servo
is initialized (e.g., at link-up on the wr node).

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

wr_ops->enable_timing_output checks the current status and only acts
on hardware when the pps must be changed (turned on or off).  Thus,
all internal uses must refer to this function, or we'll get out of
sync between actually pps activation and internall-known state.

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