Skip to content
GitLab
  1. 29 Nov, 2022 1 commit
  3. 23 Nov, 2022 1 commit
  5. 10 Nov, 2022 1 commit
  7. 02 Nov, 2022 2 commits
  9. 01 Nov, 2022 1 commit
  11. 31 Oct, 2022 1 commit
  13. 18 Oct, 2022 1 commit
  15. 21 Sep, 2022 1 commit
    • Ariel D'Alessandro's avatar
      image-uefi: Add cmdline to use SOF firmware signed with production key · a22fb939
      Ariel D'Alessandro authored 
      

The UP Squared 6000 board fails to bring up audio as it tries to load
the SOF firmware that is signed with the community key.

Starting from Linux v5.16, the following patch is included:

  commit 405e52f412b85b581899f5e1b82d25a7c8959d89 (bad)
  Author: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>

      ASoC: SOF: sof-pci-dev: use community key on all Up boards
  ---
   sound/soc/sof/sof-pci-dev.c | 10 +---------
   1 file changed, 1 insertion(+), 9 deletions(-)

Although the community key is the preferred option to allow developers
to sign their own firmware, the Skylake found in the UP Squared 6000
board only supports firmware signed with the production key.

Fix this by adding the module option (snd-sof-pci.fw_path="intel/sof")
to the kernel cmdline arguments, so the right firmware is picked up.

Signed-off-by: Ariel D'Alessandro's avatarAriel D'Alessandro <ariel.dalessandro@collabora.com>
      a22fb939
  17. 19 Sep, 2022 1 commit
  19. 15 Sep, 2022 1 commit
    • Ariel D'Alessandro's avatar
      rpi64: Copy DTBs from u-boot to firmware partition · c660dccb
      Ariel D'Alessandro authored 
      

The RPi loader tries to load a dtb for u-boot from the firmware
partition. Currently, the apertis-image-recipes recipes for RPi are
copying these dtb files from the kernel.

It makes more sense to copy these dtb blobs from u-boot rather than the
kernel, as the bootloader is the one using them. Moreover, these dtbs
are built supporting overlays, which we might use for the CANOPi
auto-detection support.

Signed-off-by: Ariel D'Alessandro's avatarAriel D'Alessandro <ariel.dalessandro@collabora.com>
      c660dccb
  21. 08 Sep, 2022 1 commit
  23. 31 Aug, 2022 3 commits
    • Edmund Smith's avatar
      Rename the template files · 4055ff54
      Edmund Smith authored and Sjoerd Simons's avatar Sjoerd Simons committed 
      In the previous commit we switched from using "submit" and "monitor"
to using "generate" and "run". This moves the template files to fit
that naming pattern too.
      4055ff54
    • Edmund Smith's avatar
      Rename the build stages to be consistent · 9c15dd08
      Edmund Smith authored and Sjoerd Simons's avatar Sjoerd Simons committed 
      We're no longer submitting jobs in the submit step, nor are we
monitoring jobs in the monitor step. This is only a separate commit
from the actual technical changes to make the earlier commit easier
to follow. This is just mechanical renaming of:

- submit -> generate
- monitor -> run

The actual files will be renamed in the next commit to make the
diffs cleaner and easier to follow.
      9c15dd08
    • Edmund Smith's avatar
      Switch to the new pattern using the lava runner · 9fd9a41e
      Edmund Smith authored and Sjoerd Simons's avatar Sjoerd Simons committed 
      This matches the changes to the apertis-test-cases repo. Previously we
had a two stage process for Lava tests of:

- generate and submit
- monitor

for each set of test cases. This is a restrictive setup, because it is
impossible to use CI_JOB_TOKEN within the lava jobs themselves
(because the token expires on completion of the submission job, before
the lava job itself may even have started).

The new model is simply:

- generate tests and pipeline
- Run each test in its own job in the generated pipeline

That means we still have pairs of jobs at the top level, one job to
generate the tests, as before, which now also generates a pipeline;
the second job is a trigger job which executes the pipeline, bringing
in the generated tests.
      9fd9a41e
  25. 24 Aug, 2022 5 commits
  27. 23 Aug, 2022 1 commit
  29. 16 Aug, 2022 1 commit
  31. 15 Aug, 2022 1 commit
  33. 04 Aug, 2022 2 commits
  35. 29 Jul, 2022 1 commit
  37. 22 Jul, 2022 1 commit
  39. 19 Jul, 2022 1 commit
    • Ariel D'Alessandro's avatar
      ostree: Prevent error on unavailable console · aac9ebc7
      Ariel D'Alessandro authored 
      

Initramfs script `local-bottom/ostree` is calling the ash built-in
command `cd -`, which tries to print the target directory to stdout. In
the case the console is still unavailable, this command fails exitting
the ostree script early.

The following strace commands were run inside the initramfs ostree
local-bottom script:

```
strace sh -c "cd / ; cd - >> /dev/null"
[...]
chdir("/")                              = 0
openat(AT_FDCWD, "/dev/null", O_WRONLY|O_CREAT|O_APPEND, 0666) = 3
fcntl(1, F_DUPFD_CLOEXEC, 10)           = 10
dup2(3, 1)                              = 1
close(3)                                = 0
wait4(-1, 0x7ffc316043cc, WNOHANG, NULL) = -1 ECHILD (No child processes)
chdir("/")                              = 0
newfstatat(1, "", {st_mode=S_IFCHR|0666, st_rdev=makedev(0x1, 0x3), ...}, AT_EMPTY_PATH) = 0
ioctl(1, TCGETS, 0x7ffc31603a90)        = -1 ENOTTY (Inappropriate ioctl for device)
write(1, "/\n", 2)                      = 2
dup2(10, 1)                             = 1
close(10)                               = 0
getpid()                                = 219
exit_group(0)                           = ?
+++ exited with 0 +++
```

```
strace sh -c "cd / ; cd -"
[...]
chdir("/")                              = 0
wait4(-1, 0x7fff4d5351bc, WNOHANG, NULL) = -1 ECHILD (No child processes)
chdir("/")                              = 0
newfstatat(1, "", {st_mode=S_IFCHR|0600, st_rdev=makedev(0x5, 0x1), ...}, AT_EMPTY_PATH) = 0
ioctl(1, TCGETS, 0x7fff4d534880)        = -1 EIO (Input/output error)
write(1, "/\n", 2)                      = -1 EIO (Input/output error)
getpid()                                = 223
exit_group(1)                           = ?
+++ exited with 1 +++
```

Note that other calls to the ash built-in `cd` command aren't an issue,
as these are not trying to print anything to stdout, so they don't fail.
So let's explicitly set the target directory, instead of calling `cd -`.

Another option it's to redirect `cd -` output to /dev/null as that's not
really useful for logging/debugging anyway.

Signed-off-by: Ariel D'Alessandro's avatarAriel D'Alessandro <ariel.dalessandro@collabora.com>
      aac9ebc7
  41. 07 Jul, 2022 2 commits
    • Walter Lozano's avatar
      Move setup_image_version to later in image build · 407adb4c
      Walter Lozano authored and Andre Moreira Magalhaes's avatar Andre Moreira Magalhaes committed 
      

The script setup_image_version.sh is used to add additional metadata to
/etc/os-relase which is a file provided by base-files package. Later during
the build additional repositories -updates and -security are enabled and
packages are upgraded, which can cause this file to be overwritten is
a new version of base-files is found.

To avoid the problem of loosing the additional metadata move the call
to the script to the end of the recipe.

It is important to note that a later update of the packages can potentially
also overwrite this file, however, in that case the additional metadata
information will be not longer be valid, so there is no real issue.

Signed-off-by: Walter Lozano's avatarWalter Lozano <walter.lozano@collabora.com>
      407adb4c
    • Ritesh Raj Sarraf's avatar
      Drop clutter related packages · cf591e95
      Ritesh Raj Sarraf authored 
      

Now that, as part of T8881, we've dropped all clutter related packages,
this MR drops the same from the image recipes

Signed-off-by: Ritesh Raj Sarraf's avatarRitesh Raj Sarraf <ritesh.sarraf@collabora.com>
      cf591e95
  43. 29 Jun, 2022 1 commit
  45. 28 Jun, 2022 1 commit
  47. 24 Jun, 2022 3 commits
  49. 22 Jun, 2022 1 commit
  51. 07 Jun, 2022 1 commit
  53. 28 Apr, 2022 1 commit
  55. 21 Apr, 2022 1 commit
    • Ryan Gonzalez's avatar
      Properly resolve full license information for Rust binaries · e5a81d3f
      Ryan Gonzalez authored 
      There are two primary issues when applying the current mechanisms for
tracking source file licensing information to Rust binaries:

- They statically link in their crate dependencies, meaning that we
  need to manually include the licenses of those crates
- The compilation unit paths only include the primary "entry point"
  into the crate, and they're a mix of absolute and relative paths

In order to resolve this:

- pkg/debhelper!18 will
  gather the license information for any source files referenced that
  are part of other packages. We can read this in here and use it to
  search the copyright information for the source files part of other
  crates.
- Specifically when encountering Rust crates, combine all the licenses
  for source files within the same source directory as the "entry point"
  file, which should generally ensure that all the source files needed
  are handled.

As the amount of information that needs to be held per package has
increased drastically, a new PackageLicenseInfo class was introduced,
encapsulating the primary copyright, external copyrights, external
sources index, and the bin2sources data.

https://phabricator.apertis.org/T8554

This should also, specifically for Rust, resolve the spin-off issues:

https://phabricator.apertis.org/T8776
https://phabricator.apertis.org/T8777



Signed-off-by: Ryan Gonzalez's avatarRyan Gonzalez <ryan.gonzalez@collabora.com>
      e5a81d3f
  57. 20 Apr, 2022 1 commit
Apertis Website Terms of Use Privacy Policy