GitLab

   129 |   TRACE("Partition differ : %s(%lu) <--> %s(%lu)",
       |                                ~~^
       |                                  |
       |                                  long unsigned int
       |                                %llu
Signed-off-by: Stefano Babic <sbabic@denx.de>

Signed-off-by: Stefano Babic <sbabic@denx.de>

strlcpy() does not exist in Linux and the function in SWUpdate was taken
from FreeBSD code. Add the missing Copyright of the author.
Signed-off-by: Stefano Babic <sbabic@denx.de>
Reported-by: Bastian Germann <bage@linutronix.de>

Signed-off-by: Stefano Babic <sbabic@denx.de>

lua just checks for .so extensions in /usr/lib/lua/<lua version>, and
without the link the library is not found. Instead of patching later
rootfs, let SWUpdate to set the link correctly.
Signed-off-by: Stefano Babic <sbabic@denx.de>

Try to create destination directory when -o parameter is given.
Signed-off-by: Stefano Babic <sbabic@denx.de>

Switch to condition variable to synchronize when an update is finished
because using mutex as semaphore raises issues with coverity tool.
Signed-off-by: Stefano Babic <sbabic@denx.de>

It is not required anymore to handle the result of the update via local
file after merging the into the generic path.
Signed-off-by: Stefano Babic <sbabic@denx.de>

This function is used just in one place to set the transaction flag. It
was thought to be generic, but the state is already saved using
set_/get_state, and this function is just a wrapper for bootloader
accessors, doing also some dangerous type conversion. This patch cleans
this up and calls directly the functions to access the bootloader's
environment.
Signed-off-by: Stefano Babic <sbabic@denx.de>

There are currently two separate paths for installing, via network
thread (used by OTA / IPC library) and from a local file. This increases
efforts during regression tests because both paths must be tested. In
case of local path, no network thread is created and file is scanned via
seek(), adding an implicit "installed-directly" to all artifacts.
Nevertheless, it happened to find dicrepancies between OTA and local
file update.
This patch is a big cleanup and drops the path from local file. The
network thread is always generated, andcode from swupdate-client is
integrated into SWUpdate to use the generic update path.
Specific functions duplicated for local path are dropped, and it is not
requested to forward in all steps if an updated is coming from local
file or from OTA.

These changes are thought to be compatible with the past, and updating
via -i (local file) is still supported. Indeed, the following different
behaviors are introduced:

- SWU check via -c is the same as dry-run mode. Not just sw-description
is parsed, but the whole update is checked without installing. This
extends the check's functionality and verifies the whole SWU.
- before these changes, -i results in an implicit "installed-drirectly"
for all artifacts. This is not maintained, and users should add the
attribute to own sw-description to avoid temporary copies on tmpdir.
Signed-off-by: Stefano Babic <sbabic@denx.de>

- Add SPDX header
- Add missing void parameter type to function declarations
- Use internal linkage for local functions (add static)
- Remove second declaration of a variable inside sub block
  (variable is shadowed)
Signed-off-by: Stefano Babic <sbabic@denx.de>

SWUpdate's support for legacy website was dropped, removed dead code.
Signed-off-by: Stefano Babic <sbabic@denx.de>

Signed-off-by: Stefano Babic <sbabic@denx.de>

Drop duplicate stuff that are already shown in the documentation
and try to make a cleaner overview of the project.
Resize logo that takes most of the place on introduction page.
Signed-off-by: Stefano Babic <sbabic@denx.de>

Signed-off-by: Stefano Babic <sbabic@denx.de>

Signed-off-by: Mark Jonas <toertel@gmail.com>

Recursive call of mkpath() may explicitly ignore the return value.
Signed-off-by: Mark Jonas <toertel@gmail.com>
Acked-by: Stefano Babic <sbabic@denx.de>

Signed-off-by: Stefano Babic <sbabic@denx.de>

Signed-off-by: Stefano Babic <sbabic@denx.de>

* Introduce IPC call GET_UPDATE_STATE.
* Use plain Linux-style return values from function and keep things simple.
* Adjust tests for Suricatta.
Signed-off-by: Sava Jakovljev <sava.jakovljev@teufel.de>

Signed-off-by: Sava Jakovljev <sava.jakovljev@teufel.de>

Signed-off-by: Sava Jakovljev <sava.jakovljev@teufel.de>

Signed-off-by: Sava Jakovljev <sava.jakovljev@teufel.de>

Signed-off-by: Sava Jakovljev <sava.jakovljev@teufel.de>
Reviewed-by: Stefano Babic <sbabic@denx.de>

Signed-off-by: Sava Jakovljev <sava.jakovljev@teufel.de>

Calling copyfile the parameter "encrypted" is used as bool and not as
int. Replace type in function and where it is called.
Signed-off-by: Stefano Babic <sbabic@denx.de>

Signed-off-by: Sava Jakovljev <sava.jakovljev@teufel.de>

install_remote_swu() will return a good state in case of two error states.
Fix them to return FAILURE like most of the other errors.
Signed-off-by: Bastian Germann <bage@linutronix.de>
Acked-by: Stefano Babic <sbabic@denx.de>

TEXTRANGE_TO_STR is called with uri.portText which might contain
{NULL,NULL} if there is no port information in a swuforward URL.
In that case, evaluate to NULL instead of calling substring.
Signed-off-by: Bastian Germann <bage@linutronix.de>
Acked-by: Stefano Babic <sbabic@denx.de>

wc_AesCbcDecrypt does not update the initialization vector for subsequent
calls to it. So the decryption for each buffer but the first one results
in the first AES block being garbage.

Call wc_AesSetIV with the last encrypted block after each wc_AesCbcDecrypt.
Signed-off-by: Bastian Germann <bage@linutronix.de>
Acked-by: Stefano Babic <sbabic@denx.de>

The additional buffer in the update step is not needed.
Make the decryption operate on the buffer that is given to the function.
Signed-off-by: Bastian Germann <bage@linutronix.de>
Acked-by: Stefano Babic <sbabic@denx.de>

Termination string for colors was wrongly set to '0' instead of '\0',
resulting in garbage in log files.
Signed-off-by: Stefano Babic <sbabic@denx.de>

Signed-off-by: Stefano Babic <sbabic@denx.de>

Last commit add an additional call to save_state()
Signed-off-by: Stefano Babic <sbabic@denx.de>

Signed-off-by: Sava Jakovljev <sava.jakovljev@teufel.de>

This removes file like built-in.o.tmp that can break build on multicore
systems.
Signed-off-by: Stefano Babic <sbabic@denx.de>

In some use cases, even sw-description is required to be encrypted. This
is useful in case sw-description contains Lua code and it is not
desirable to expose this code.

A new CONFIG_ENCRYPTED_SW_DESCRIPTION is added - this is
at compile time because it is a security option.
Signed-off-by: Stefano Babic <sbabic@denx.de>

The function `get_value_libconfig` contains an incorrect type cast
when parsing booleans which results in overriding other memory.

The bug can be triggered by having an sw-description like this:

```
bootloader_transaction_marker = false;

[...]

foo = {
    bar: {
        [...]
    }
};
```

and using SWUpdate's selection mechanism (`-e foo,bar`).

The selection mechanism will *not* work when sw-description contains
`bootloader_transaction_marker`.

The reason is that the value of `bootloader_transaction_marker` is
stored in the struct `swupdate_cfg` which contains the fields:

```
    bool bootloader_transaction_marker;
    char software_set[SWUPDATE_GENERAL_STRING_SIZE];
```

The incorrect type cast (`int` instead of `bool`) results in overriding
the value stored in the `software_set` and consequently breaks the
software selection mechanism.
Signed-off-by: Michael Adler <michael.adler@siemens.com>
Signed-off-by: Christian Storm <christian.storm@siemens.com>
Reviewed-by: Stefano Babic <sbabic@denx.de>

Signed-off-by: Christian Storm <christian.storm@siemens.com>
Signed-off-by: Stefano Babic <sbabic@denx.de>

Signed-off-by: Christian Storm <christian.storm@siemens.com>