libdt: support "partitions" subnode on mtd devices ...

In case there is an additional "partitions" subnode between the device
itself and its partitions like /soc/spi@?/mram@?/partitions/partition@?
also try the parent of the parent before giving up finding the device
node.

See kernel docs for reference:
  Documentation/devicetree/bindings/mtd/partition.txt

Signed-off-by: Bastian Stender <bst@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

rework logging functions ...

We already use pr_* and dev_* functions. Replace some more
fprintf(stderr,...) with appropriate pr_* function calls. Also
provide control over which urgency level is printed.

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

of_print_property: fix compiler warning ...

src/libdt.c: In function ‘of_print_property’:
src/libdt.c:97:13: warning: format ‘%s’ expects argument of type ‘char *’, but argument 2 has type ‘const void *’ [-Wformat=]

Calm the compiler with an explicit cast.

of_get_devicepath: correct comment and add link to external patch

of_get_devicepath: make partition finding more robust ...

If a backing device for a partition is neither an eeprom nor an mtd nor a
block device I don't know what it should be. Still fall back to the generic
case where the partition doesn't have a related of_node.

of_get_devicepath: if the device is an mtd udev_parse_mtd is all we need. ...

So the function should return success and not fall back to the "there is no
associated device" case.

Fixes: d83976b8163d ("libdt: add block devicepath extraction")

of_get_devicepath: simplify code flow ...

This reduces indention level and makes the code more easy to follow.

There is no semantic change (but now it's obvious there is a bug which
is fixed in the next commit).

Fixes: d83976b8163d ("libdt: add block devicepath extraction")

remove two unused variables ...

Fixes: 53df62ae2fe3 ("state: describe backend with phandle")

libdt: take the parent to search for block devices ...

Signed-off-by: Michael Grzeschik <m.grzeschik@pengutronix.de>

libdt: only enumerate devices we are interested in ...

Only devices of certain subsystems can have device nodes. Limit
the enumeration to these subsystems, namely platform, i2c, spi
and mtd. pci maybe missing here, but let's see once we are there.

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

libdt: cache of_find_device_by_node_path results ...

Scanning udev for devices with a OF_FULLNAME property can take
a long time. Since we may do this multiple times cache the results
during first run to speed up subsequent runs.

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

libdt: add block devicepath extraction ...

This allows barebox state to find the backend node on block
devices.

This is a reworked version of the patch. The outdated patch
did not specify the parent correctly resulting in an
incorrect subtree. This lead to the wrong block device
being returned.

Signed-off-by: Bastian Stender <bst@pengutronix.de>

Revert "libdt: add block devicepath extraction" ...

This reverts commit 6e91248.
It currently breaks barebox-state. /dev/loop0 is always selected.

libdt: add block devicepath extraction ...

This allows barebox state to find the backend node on block devices.

Signed-off-by: Bastian Stender <bst@pengutronix.de>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

libdt: fix detection of eeprom devicepath ...

If nodepath can directly be translated to a device, the device can also
be a eeprom. Set default size to 0. The real size can be got from the
path.

Signed-off-by: Jan Remmet <j.remmet@phytec.de>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

libdt: add missing newline in error message ...

Reported-by: From: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

libdt: of_read_devicetree(): try to read fdt from /sys/firmware/fdt first ...

Which is faster than walking over the filesystem representation of the device
tree in /proc or /sys.

Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

libdt: fix printf related bugs ...

src/libdt.c:97:11: warning: format string is not a string literal (potentially insecure) [-Wformat-security]
                        printf(data);
                               ^~~~
src/libdt.c:2241:5: warning: data argument not used by format string [-Wformat-extra-args]
                                node->full_name);
                                ^
src/libdt.c:2248:54: warning: more '%' conversions than data arguments [-Wformat]
                fprintf(stderr, "%s: no 'label' property found in %s\n", partition_node->full_name);
                                                                  ~^

Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>

state: describe backend with phandle ...

We used to describe the backend with a combination of a nodepath
and partname:<name>. This is cumbersome and error prone to parse.
Instead, describe the backend with a phandle that directly points
to the partition. The resulting code is easier to read and the
devicetrees are easier to understand. Tested with both mtd and
EEPROM device, both with partitions with an offset != 0.

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

scan_proc_dir: Add missing phandle initialization ...

node->phandle was not set when /proc/device-tree was read.

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

Add some comments ...

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

remove dos partition parser ...

Instead of matching DOS partitions with partname:<number> this
should be made explicit with a different type parser, like dospartid:<id>.
Since this is unused just remove it for now.

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

document of_parse_partition_from_path ...

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

libdt: fix whitespace damage ...

Signed-off-by: Steffen Trumtrar <s.trumtrar@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

libdt: add sysattr for partition ...

mmc devices only have a "name" property that reads 00000.
The partitions have a property "partition" telling the
partition index.

Signed-off-by: Steffen Trumtrar <s.trumtrar@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

libdt: refactor partition parsing ...

Signed-off-by: Steffen Trumtrar <s.trumtrar@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

libdt: search for devnode if partition not found ...

Signed-off-by: Steffen Trumtrar <s.trumtrar@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

libdt: add function to search the devicenode in subdirs ...

Signed-off-by: Steffen Trumtrar <s.trumtrar@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

libdt: fix segfault ...

strdup will segfault, if udev_device_get_devnode does not find the part.

Signed-off-by: Steffen Trumtrar <s.trumtrar@pengutronix.de>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>

2nd commit ...

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>