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

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

Also fix makefiles and documentation
Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

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

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

This allows other process to access port information without
making several RPC calls.
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 stdout and stderr of the basic WR tools are piled to the "logger"
process, to get actual logs saved, if so configured for.

Thus, we should not reopen the output channels to /dev/null
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This is the first step to move the "get_port_state" query to shared
memory.  The internal structures of hal_ports.c will be allocated
in shared memory, so the calls to halexp_* will just convert the data,
without context-switching to the hal process.

I couldn't add the functions to libwr/hal_client.c where they belong,
because the choice of using the same "halexp" names for local and remote
calls would lead to duplicate symbol definitions when linking the hal
(the hal process links libwr, but can't link hal_client.o).
Thus, libwr/hal_shmem.c is created, to host the parts of hal/hal_ports.c
that are needed to convert data.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Acked-by: Maciej Lipinski <maciej.lipinski@cern.ch>

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

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

LOCK_STATE_* macros were defined locally and not used
(only LOCK_STATE_NONE was used, but 0 is better since the
relevant function says "returns 0 or 1").
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

HAL_PORT_MODE_* were defined locally and never used. The code
uses HEXP_PORT_MODE_* instead (defined in <hal/hal_exports.h> and
used in several places).
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

After this commit, all symbols in hal_ports are of the form hal_port_*
(all but one, that is not used and I'll remove it in another commit).
I'm not smart enough to track the origin of the various names, so I
need this.  I need this especially in this file because I'm going to
export stuff using shmem, to avoid a lot of context switches.

Thus, the "private" structure hal_port_state is not a typedef any
more, as a first step in becoming public. It doesn't make sense to
have two almost-identical structures, where one is copied to the other
one item at a time to deal with the minor differences. (It happened
for some reason: I'm not complaining hard, but it's time to get rid of
this).

Unfortunately, two functions were called "halexp", and relied on a
prototype shared by the RPC server (this file) and the RPC client.
While the convention made sense at first, it now shows its limits, so
the exported one is called hal_port_* like everything in the file, but
for this I need to add structure names (but I left the typedef at this
point, to avoid massive name changes around).

If I didn't introduce bugs, the code is the same as it was. I did it
one symbol at a time while build-checking each steps (and I even have
the commits to track any error).
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

pps_gen.c had a missing semicolon because of this bug. Fix the
macro to behave properly withing if/else. And fix the user, adding
the missing semicolon.

We all know that these two functions should be static inline, and the
fpga base pointer should be void *, but that's for later.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This header is almost obsolete, as we'll move to SDB soon.
Meanwhile, sort out the tab/space mess and indent with spaces only,
so it looks good to everybody (including greg who insists on one
tab every 2 spaces -- which works for indentation, but not for
for alignment at end-of-line).
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

We'd better use less CPU power to run the HAL, even if this shows down
response times for management queries (until we move most of it to
shmem).
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

The code was hardwired to /dev/ttyS2 and 115200 baud. So we are sure
nobody is using it, and it actually can't be used at all.

If needed, we know it's in the history.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Ackec-by: Grzegorz Daniluk <grzegorz.daniluk@cern.ch>

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

The compiler noted already, by not generating code.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

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

I didn't take all suggestions by Lindent, I avoided to commit
the stupid ones (when on separate hunks).

This mainly changes the spacing of "if( condition)", spaces after
commas in arguments, the placing of the pointer symbol and
the placement of braces.  I would have avoided the last item, but
it's not simple.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This is a semi-automatic pass. To work on this I need to have it
in 80 columns, and sanely indented. This pass is not changing
brace placement or whatever: it is just a "fill-region" in emacs plus
manual split of strings and shortening of comments. You can
"git diff --word-diff=color" to check.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This allows private headers to be clearly private (in userspace/wrsw_hal/).
rt_ipc.h is used by phytool too, so it is not private and is now in include/.

As a side effect, remove -I../wrs_hal from makefile.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

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

To delay for less than 1ms (was: 10ms) we must busy-loop, because the
kernel is jiffy-based. No way out of it.

The implementation, however, was based on gettimeofday(), which means
we couldn't really strace the hal process, as it appeared to
only call gettimeofday all the time.

This code uses a local software loop, after a calibration step of
around 20ms (depends on CPU speed). Moreover, by calling usleep() for
longer-than-1ms delays it doesn't use 100% cpu power during the
initial led blinking.

Now strace on the process shows what it really is doing (besides
user-spac-only i2c accesses).
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

We don't need to print binary and get mad with static storage.
If any, we print hex.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

shw_malloc and shw_free are not used (well, i2c_free used shw_free,
but there is no associated alloc anywhere, so I removed it too).

The idea of a "malloc-or-exit" kind of function is good, but we'd
better make sense of things (and simplify where unused legacy remains)
before finding and fixing all unlikely error paths.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This is a completely automatic pass, before we start making order
in this library (note: there is a missing semicolon in pps_gen.c:82
but it will be fixed in a later commit -- the bug is in the macro
being called, so the missing semicolon is ok for the compiler).
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

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

This is the new IPC mechanism (not RPC) to pass port status and
other status-like information among processes, to avoid lenghty RPC
calls.

Each WR process has 32k of shared space.

Thanks to Adam for some fixes.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

This changes the kernel configuration, in two ways:

1- It uses CONFIG_NOHZ, which enables Hight Resolution Timers

Unfortunately this has no effect, as the timeevent device for AT91
doesn't support oneshot mode.  This means that if you usleep(100)
you'll wait 10ms.  Thus, we can't remove the busy-looping in libwr/util.c .

This is the output of my libstamp sample program:

     0.003 -- stamping overhead
    12.243 -- usleep     1 usec
    10.005 -- usleep     1 usec
     9.963 -- usleep     2 usec
    10.037 -- usleep     5 usec
    10.083 -- usleep    10 usec
     9.905 -- usleep   100 usec
    10.394 -- usleep  1000 usec
    19.775 -- usleep 10000 usec
   109.925 -- usleep 100000 usec

2- It moves CONFIG_HZ from 100 to 1000. Thus, we have 1000 timer interrupts
per second.

This means, in practice, that if you usleep(100) you wait 1ms.  This
costs around 2% of CPU power (measured by running a cpu-busy program
before and after the change, in both cases after killing all
processes), but makes stuff millisecond-grained instead of
10ms-grained. This is a benefit for the PTP daemon and everything uses
timeouts in system calls.

This is libstamp again:

     0.003 -- stamping overhead
     0.662 -- usleep     1 usec
     0.869 -- usleep     1 usec
     0.976 -- usleep     2 usec
     0.951 -- usleep     5 usec
     0.993 -- usleep    10 usec
     0.999 -- usleep   100 usec
     1.999 -- usleep  1000 usec
    11.006 -- usleep 10000 usec
   101.043 -- usleep 100000 usec
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This updates binaries/wrboot-* to be more commented and complete.
It also adds wrboot-install and wrboot-nand, that I personally used.

As a side effect, "mem=64m" is removed by all command lines, as it has
been properly autodetected for a few years now, so it is not needed
any more.

Documentation is updated too.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

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

Also, fix Makefiles accordingly.
Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

This commit makes no changes, as shown by "git diff -w" before
committing, but blank spaces at end-of-line make my editor scream
(because I'm pedanting in not pushing those to the kernel).
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>