nvvar: Fix creation without value ...

When a new nvvar is added without a value we want to get the value
from the corresponding globalvar. We do this by using nv_set which
ends up freeing the exact string that we passed into nv_set as
value. This results in a corrupt value. Fix this by assigning a
value manually without calling nv_set.

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

globalvar: Fix compiler warning ...

Fixes:
common/globalvar.c:393:7: warning: 'pname' may be used uninitialized in this function [-Wmaybe-uninitialized]

This is a false positive, pname cannot be uninitialized. Silence this
warning.

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

globalvar: Make locally used function static ...

Fixes:

common/globalvar.c:382:5: warning: no previous prototype for 'globalvar_simple_set' [-Wmissing-prototypes]

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

complete: Add completion for nv and globalvar commands ...

The 'nv' command is often used to create a nv variable
for an existing global variable, so add a command completion
function for this.

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

globalvar: introduce globalvar_add_simple_bitmask ...

Using the just introduced param_bitmask this adds the corresponding
globalvar convenience function.

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

globalvar: Allow full variable name in globalvar_add ...

As a convenience for users allow to pass the full name, including
the leading "global.", to globalvar_add().

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

nv: Allow full variable name in nvvar_add ...

As a convenience for users allow to pass the full name, including
the leading "nv.", to nvvar_add().

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

globalvar: Also create globalvars from for nonvolatile device vars ...

nv variables beginning with "nv.dev.<devname>.<varname>" are directly
mirrored to <devname>.<varname> and there is no globalvar for it. To
make it a bit more consistent and to increase the visibility of the
nonvolatile device variables create globalvars for them aswell.

With this the propagation flow has changed from:

nv.dev.<devname>.<varname>
  -> <devname>.<varname>

to:

nv.dev.<devname>.<varname>
  -> global.dev.<devname>.<varname>
    -> <devname>.<varname>

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

nv: simplify nvvar_add ...

We do not need to have an extra code path when the variable already
exists, instead setting an existing variable can be done in the
variable creation code path aswell.

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

globalvar: Allow to remove multiple globalvars using wildcards ...

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

globalvar: sync with nvvars ...

This patch fixes the behaviour when a driver creates a globalvar using
globalvar_add_simple_[string|int|bool|enum|ip]) *after* nvvars are
initialized and this globalvar is overwritten with a nvvar. Currently
this fix is not needed because all globalvars are registered before the
nvvars are initialized.

We have two different typed of globalvars. The first type, here referred
to as qualified globalvars, has a backend variable storage (the ones
created with globalvar_add_simple_[string|int|bool|enum|ip]), the other
created with globalvar_add_simple only has a dynamically allocted string
as backend.

Normally during startup of barebox the qualified globalvars are
registered and during load of nvvars are synced with the values from the
nvvars. Everything works fine in this case. However, when during nvvar
initialisation a globalvar for a nvvar does not exist, then it is
registered as unqualified globalvar. When then later some driver wants
to register a qualified globalvar for which a unqualified globalvar
already exists, it will get a -EEXIST. This is not the expected
behaviour. Instead, the current unqualified globalvar should be removed,
recreated as qualified globalvar and then afterwards synced with the
corresponding nvvar. This behaviour is fixed with this patch.

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

globalvar: Move static inline functions to common/ ...

These functions will get bigger in the next patch which disqualifies
them as static inline functions.

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

nv: Fix variable removal in nvvar_save() ...

When nv variables are removed during runtime then they are
present again when saved with nvvar_save(). This is because nvvar_save()
does not delete variables that exist on the saved environment. Delete
/nv on the saved environment before saving the new variables.

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

nv: Allow wildcards when removing NV vars ...

With this patch 'nv -r' can also take "*" and "?" wildcards for nv
variables. This makes it easier to remove multiple nv variables.

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

nv: Use dev_remove_param to delete nv variable ...

dev_remove_param() is exactly for the purpose of removing a device
parameter, so use this function instead of open coding the
functionality.

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

nv: Save nv variables on shutdown ...

With this patch nv variables are automatically saved whenever barebox
shuts down (that is 'reset' is executed or a kernel is started). With
this the additional 'saveenv' step becomes unnecessary.

The nv variables are stored in the environment and the estasblished
behaviour is that files in the environment must be manually saved
using 'saveenv'. This behaviour shall be kept for now, so this patch
cannot just call 'saveenv' since that would save the modified
environment files aswell. Instead we read the environment from the
device, modifiy the nv variables and save the environment back.

Since this changes a long existing behaviour messages are printed the
first time a nv variable is modified and during shutdown when the
variables are actually saved.

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

nv: Do not save nv variables while loading ...

When reading nv variables from the storage in /env/nv we do
not need to write back the value to the file we just read from.
Optimize this a bit and make it unnecessary.

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

Introduce non volatile device variables ...

Non volatile device variables are used to make device parameters
persistent. They are like normal non volatile variables, but set
the values of the device parameters with the corresponding name.
Every nv variable beginning with nv.dev is a non volatile device
variable. They have the form nv.dev.<devname>.<paramname> and act
on the parameter <paramname> of the device named <devname>.

The non volatile device variables are designated for example for
video modes, ethernet device ip addresses or mtd partitioning.

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

nvvar: Simplify by using dev_set_param() ...

The combination of get_param_by_name()/param->set() can be replaced
by using dev_set_param()

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

nvvar: Simplify by using nv_set() ...

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

Merge branch 'for-next/misc'

Merge branch 'for-next/include-cleanup'

globalvar: Allow to remove globalvars ...

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

Kconfig: Create Kconfig symbol for NVVAR ...

nvvar support not only needs globalvar, but also persistent
environment storage. Add a separate default-y option which
depends on ENV_HANDLING for this case. Make the option visible
to let the user decide whether he wants to have this option
and add a help text to make this decision easier.

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

stdio: Replace FILE functions with filedescriptor functions ...

We have defined stdin, stdout and stderr as integer file descriptors,
but normally they should be FILE *. Also fprintf, fputc and fputs take
file descriptors instead of FILE *. As FILE * are inconvenient in the
barebox environment replace the f* functions with the corresponding d*
functions. dprintf is POSIX conform whereas dputc and dputs are barebox
specific, but do not conflict with any stdc function. fgetc is unused
and can be removed without replacing it.

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

string: Fix (v)asprintf prototypes ...

Our asprintf and vasprintf have different prototypes than the glibc
functions. This causes trouble when we want to share barebox code
with userspace code. Change the prototypes for (v)asprintf to match
the glibc prototypes. Since the current (v)asprintf are convenient
to use change the existing functions to b(v)asprintf.

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

globalvar: Add support for printing all global variables ...

This could previously only be done with 'devinfo global'. While
this is still possible this adds a more direct access via the
globalvar command. This variant also adds a '*' in front of
the variable if the corresponding non volatile variable exists.

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

Add support for non volatile variables ...

This adds (back) support for non volatile variables. Non volatile
variables are variables which are stored in the environment over
reboot. They are used in the same way as the global variables, but
with a 'nv' command and device. The variables are stored under
/env/nv/, one variable per file. Adding a nv variable automatically
adds a global variable with the same name. Changing a nv variable
also changes the same global variable, but not the other way round.
This allows for example to configure the username as:

nv user=sha; saveenv

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

param: let dev_add_param return the newly created param ...

dev_add_param creates a new parameter so it makes more sense to
return it than to return an error code. Since the return value
is hardly ever checked this is only a small patch.

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

treewide: Add missing includes ...

A lot of files rely on include/driver.h including include/of.h (and
this including include/errno.h. include the files explicitly so we can
eventually get rid of including of.h from driver.h

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