Buildroot forces procps-ng to install binaries in the /bin insted of /usr/bin.
As a side effect libprocps.pc was installed at /lib/pkgconfig instead of
/usr/lib/pkgconfig. Which directory is not included into final image.

Added patch taken from buildroot's repo:
0e07a5ede package/procps-ng: install libprocps.pc in staging/usr/lib/pkgconfig/

Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

Fix the warning (for perl < 5.21.x) whihch becomes an error
(for perl >= 5.21.x):

defined(@array) is deprecated at kernel/timeconst.pl line 373.
  (Maybe you should just omit the defined()?)

This error pops up on newer systems. It is not related with the change of
the buildroot, but with the host version of perl.

Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

Remove binaries/usr/lib/libbfd-2.21.1.so and /usr/bin/nm from the repo.
Instead use the compiled versions in buildroot.

Patch the installation rule to install only "nm", when
PACKAGE_BINUTILS_TARGET is not used(!).
libbfd-2.21.1.so is installed by default by the installation rule of
PACKAGE_BINUTILS.

Binary nm was added in the commit:
d2a486cc

 rootfs_override: add /usr/bin/nm and its libbfd

Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

Add if-mib/ifXTable to the SNMPd. It adds 64bit (HC) counters to the ifTable.

Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

Second test of $? value was not always result of command
"tftp $eth0.ipaddr/wrboot wrboot", but sometimes result of first if statement
(in case first tftp was successful).

In other words, when there was file wrboot-$eth0.ethaddr on tftp server, script
was still fetching and executing wrboot.

Signed-off-by: Adam Wujek <adam.wujek@cern.ch>

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...

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>

Standard JEDEC ID is only 24bits to identify a DF chip.
It also has an optional Extended Device Info (EDI) on bytes 4 and/or 5
that need to be read to differentiate some DF chips. (i.e, the
difference between AT45DB641E and AT45DB642D is made by byte 4).

This patch is similar to the kernel one #66595328

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

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

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

Commit 91c22de9

 replaced some barebox patches, but I forgot to remove
the old ones, so build fails.

This fixes the problem. Moreover, there is a typo, which is
now fixed.

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

PATCH SQUASH

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

Benoit Rat authored 10 years ago and Alessandro Rubini committed 10 years ago

Standard JEDEC ID is only 24bits to identify a DF chip.
It also has an optional Extended Device Info (EDI) on bytes 4 and/or 5
that need to be read to differentiate some DF chips. (i.e, the
difference between AT45DB641E and AT45DB642D is made by byte 4).

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

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

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

Benoit Rat authored 10 years ago and Alessandro Rubini committed 10 years ago

We don't need to wait these extra 3 seconds because we now load UBIfied
which takes more time than these 3 seconds.
We also add the autoboot_select in config because it is a hidden feature
asked by diferent persons.

This bug exists since time 0 (e484c3eb

, July 2011, at least).

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

This reverts commit 352d0d93

.
There is no need to do that, things work with full memory.

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

With mem=64M we have some sdram corruption, with "bad page state".
This is unexplained, and the sam-ba code is so difficult to fix, that
we'd better just install with 32M of sdram.  At runtime we use a fixed
at91boot, so 64M work properly.

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

Grzegorz Daniluk authored 10 years ago and Alessandro Rubini committed 10 years ago

This actually has no effect, but since the item is there, we'd better
have it correct.

Grzegorz Daniluk authored 10 years ago and Alessandro Rubini committed 10 years ago

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

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

This is similar to the kernel patch, but its impact is less
important: barebox is only reading nand in our setup, and it looks
like it never failed (on the other hand, the times were definitely out
of specifications).

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

We used the timing from the g45ek, but ours is slower. Actually, ours
is slower only in 16-bit mode, and has the same timing when used in
8-bit mode.

Our timing setup was outside of the allowed specification, and resulted
in random and unreproducible failures.  The "visible" effect is
that reading 100MB takes 13.4s instead of 11.4s, measured with:

  time dd bs=1k count=102400 < /dev/mtd1 > /dev/null

You can try another, still slower, setup just by hand, like this
(in this order):

   devmem 0xffffe83c w 0x00051003
   devmem 0xffffe838 w 0x000f000f
   devmem 0xffffe830 w 0x04060406

but the current settings (41003, 000c000c, 02040204) are alread within
specification.

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

Kernel support for verifying writes is not subpag-aware, and the
net result is a complete failure.

This reverts commit bdccd4d9.

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

This also change the flash_wrs script, to match new conventions
(i.e. we use the run-time initramfs for installation, and so on)

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

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

This changes partitions, mainly adding a kernel-visible mapping of
what we ha in dataflash.  NAND changes are compatible, as only empty
space is used differently.

This also includes a new barebox.bin, that you can just write
to /dev/mtd3 after booting the new kernel.

Documentation is updated to reflect these partitions, but there also
old leftovers that are fixed in this same commit.

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

Some WRS speciments won't work correctly with automatic speed setup.
This patch forces full-speed on the device (instead of the
autodetected high-speed), and thus they work.

Speed is not a problem anyways, because it is just a serial port.

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

Without this, the switch may be unable to boot NFS-Root over a new system
(e.g. my laptop, currently linux-3.14 with full NFS support.

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

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

Since there is no env config file, envfs_check_super() call fails to find
correct magic number and makes envfs_load function to return before setting
default MAC addresses.

Again, suboptimal formatting is now an error in new makeinfo. Worse,
this looks correct, but the complain is there anyways.  Hack a fix,
killing a newline that make the parser report the error.

Hope we are done: nobody outside gnu is fool enough to use makeinfo.
Me excluded, obviously (see wr-switch-sw/doc).

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