| Age | Commit message (Collapse) | Author |
|---|
|
By omitting an explicit exit code after the recovery command we
will get the exit code of the last command. This makes sense,
since we would be able to tell whether recovery actually succeeded
or not.
Signed-off-by: Stefan Agner <stefan.agner@toradex.com>
Acked-by: Marcel Ziswiler <marcel.ziswiler@toradex.com>
|
|
Since we introduced set -e commands which fail will exit silently.
This leads to no error message when the grep which tests for
specified parameters failed.
We could solve it with || true, but it seems more lightweight to
just test for the strings inside the if statement. It also checks
for the full length (e.g. disallows "25" as RAM size).
The later fixed only 256 MB RAM size copy paste issue.
Signed-off-by: Stefan Agner <stefan.agner@toradex.com>
Signed-off-by: Marcel Ziswiler <marcel.ziswiler@toradex.com>
|
|
Unix commands usually exit successfully when help was explicitly
asked for. However, in case arguments are incomplete or otherwise
wrong, exit with error.
Signed-off-by: Stefan Agner <stefan.agner@toradex.com>
Acked-by: Marcel Ziswiler <marcel.ziswiler@toradex.com>
|
|
printf allows to print escape sequences (e.g. for colored output)
and is part of the POSIX standard.
Signed-off-by: Stefan Agner <stefan.agner@toradex.com>
Acked-by: Marcel Ziswiler <marcel.ziswiler@toradex.com>
|
|
Adjusted update scripts for fast boot mode migration as per the
following article on our developer website:
http://developer.toradex.com/knowledge-base/linux-booting#eMMC_Fast_Boot_Mode_ApalisColibri_iMX6
Signed-off-by: Marcel Ziswiler <marcel.ziswiler@toradex.com>
|
|
e.g. look for files in ${board} subdirectory rather than in the media's root.
Signed-off-by: Max Krummenacher <max.krummenacher@toradex.com>
Signed-off-by: Stefan Agner <stefan.agner@toradex.com>
|
|
Signed-off-by: Max Krummenacher <max.krummenacher@toradex.com>
Signed-off-by: Stefan Agner <stefan.agner@toradex.com>
|
|
Signed-off-by: Max Krummenacher <max.krummenacher@toradex.com>
Signed-off-by: Stefan Agner <stefan.agner@toradex.com>
|
|
Write kernel and device tree into seperate UBI Volumes. This
allows to use a lot less UBI/UBIFS support in U-Boot, which
should lower the risk of hitting bugs in this area. Boot times
are also slightly faster (measurements showed an improvement of
~150ms).
Signed-off-by: Stefan Agner <stefan.agner@toradex.com>
Acked-by: Max Krummenacher <max.krummenacher@toradex.com>
|
|
When using TFTP update, the rootfs must be in decimal numbered chunks on the
tftp server. U-Boot reads the chunks, but does count hexadecimal thus
limiting the number of chunks to 10.
Signed-off-by: Max Krummenacher <max.krummenacher@toradex.com>
Signed-off-by: Marcel Ziswiler <marcel.ziswiler@toradex.com>
|
|
The U-Boot binary for the IT temperature range changed names from
u-boot-it.imx to u-boot.imx-it.
The flash_*.scr scripts did still use the old name resulting in the U-Boot
not being found on the SD card or TFTP server.
Signed-off-by: Max Krummenacher <max.krummenacher@toradex.com>
Signed-off-by: Marcel Ziswiler <marcel.ziswiler@toradex.com>
|
|
This has been missing since our move to boot from the eMMC boot partition.
While at it clarify update.sh help. _it applies only to
Apalis iMX6 IT modules.
Signed-off-by: Max Krummenacher <max.krummenacher@toradex.com>
Acked-by: Marcel Ziswiler <marcel.ziswiler@toradex.com>
|
|
Change the name form u-boot-it.imx to u-boot.imx-it.
This prepares for the upcomming move of the i.MX 6 based modules BSP
to meta-fsl-arm-extra.
Signed-off-by: Max Krummenacher <max.krummenacher@toradex.com>
Acked-by: Marcel Ziswiler <marcel.ziswiler@toradex.com>
|
|
The command 'mmc partconf 0 1 1 0' seems to have no effect when it is in the
middle of our migration cmdline resulting in U-Boot still being loaded from
the user partition.
Putting it as the last command seems to solve the issue.
http://developer.toradex.com/software-resources/arm-family/linux/linux/release-details?view=all&issue=14774
Signed-off-by: Max Krummenacher <max.krummenacher@toradex.com>
Acked-by: Marcel Ziswiler <marcel.ziswiler@toradex.com>
|
|
With the cfgblock command in U-Boot we now have a better solution.
Signed-off-by: Max Krummenacher <max.krummenacher@toradex.com>
Acked-by: Marcel Ziswiler <marcel.ziswiler@toradex.com>
|
|
Don't write the eMMC twice on Apalis in the not migrate case.
|
|
|
|
Up until now, U-Boot for Colibri VF50/VF61 did not had an
environment variable board_name. Use the soc environment
variable to set the board_name.
|
|
To avoid code duplication call fwd_blk from fwd_mmc. This adds
one more indirection, but since flash_mmc.img is a compatibility
script anyway it is not the common case...
|
|
Source the update script only if loading succeeded. This avoids
loops if board_name is not set. Also add a test if board_name is
really set since this variable is/was not available in older
U-Boots.
|
|
Further generalise U-Boot update scripts by using ${board_name} instead
of hard coding module name like apalis_t30, colibri_t20 or colibri_t30.
|
|
Even though our downstream L4T based kernel does not use any
device tree stuff prepare for easy deployment thereof by using the
correct dtb name.
In the Ethernet aka TFTP update case this actually fixes an issue where
it previously tried to TFTP the whole ${board_name} directory which at
least on my setup failed with a permission denied which it retried
forever.
|
|
Update our U-Boot update scripts to accommodate our new unified eMMC
layout:
boot area partition 1 aka primary eMMC boot sector:
with U-Boot boot loader and the U-Boot environment before the
configblock at the end of that boot area partition
Handle migration of boot loader and configblock on Apalis/Colibri iMX6.
When doing a full update also explicitly flash kernel and FDT from the
update medium in order to allow for easy updating thereof.
|
|
Further clean-up and unify our update scripts:
- Universally mention Ethernet aka TFTP, SD card and USB flash drive as
possible update media.
- Universally mention detailed recovery/update procedures being
available on our developer website.
- Universally split usage and flashing instructions (-f).
- Make sure we don't exceed the 80 characters line length limit.
- Fix various spelling issues.
|
|
Now that both Apalis/Colibri iMX6 as well as Apalis/Colibri T30 will
use the same eMMC layout also concerning boot loader, environment and
configblock locations in the primary eMMC boot area partition update
resp. description.
|
|
The following commit:
f463f754048ebc3325025f0e1506b03cf14203a5
update.sh: use moduletype specific folder for output files
moved the output folder check above the module type selection which now
fails in case of debug aka recovery mode only operation with the
following error:
does not exist, exiting
This patch fixes this by not erring out in the debug aka recovery mode
only operation if no output folder is set.
|
|
With update.sh one creates files on e.g. a SD-card or a tftp server
which enables updating the SW on a module from U-Boot running on
said module.
This commit adds a module type specific directory on that output media
so that update data for multiple module types can be stored on one media.
Generic U-Boot scripts in the media's root directory do select the
matching update data depending on module type.
|
|
|
|
To flash a large rootfs over tftp the rootfs must be split into chunks
of 64MB on the tftp server.
Add the option '-c' to the update.sh script which does this automatically.
Use 64MB chunks on all module types.
|
|
While at it:
- remove the default directory for the -o parameter
- cleanup text and sort variables
|
|
Switch to using the temperature Chamber validated 800 MHz BCT
Apalis_T30_2GB_800Mhz.bct.
Please note that for now this is a fixed 800 MHz BCT which does not
allow for memory DVFS (e.g. 400/800 MHz switching like currently only
possible on Colibri T30).
|
|
Follow the rename imx6q-apalis-eval_v1_0.dtb to imx6q-apalis_v1_0-eval.dtb
in the kernel.
The v1_0 denotes the Apalis module version, not the evaluation board version.
|
|
While we were checking whether the mkfs.ubifs tool has the magic
--space-fixup option available we never actually enabled that option
to be used when generating our UBIFS root file system images.
Thanks Bhuvan for spotting this having actually been in our update
scripts for almost two years now!
|
|
Now that we exit on failure the test for a user accessible parted must not
return an error code.
|
|
On recent Ubuntu distributions gawk no longer supports awk -V which makes
the test fail.
Simulate our awk use case and test the result instead.
Additionally set 'set -e' to exit immediately on errors. Fix places where
a command returns not zero in 'normal' situations.
E.g. A full disk leads otherwise to funny effects.
Allow relative OUT_DIR.
|
|
The eMMC layout comments are confusing and in some regards outdated.
Clean it up.
While at it move the calculation of the rootfs file system size to
the rootfs creation to not confuse the rootfs file system size with
the size of the rootfs partition.
|
|
U-Boot now has the command cfgblock which can be used to create a valid
configblock.
Alternatively one can be created with create_configblock.sh.
|
|
Fix and enhance messages durring Colibri VF50/VF61 upgrade.
Fix missing run command
|
|
Was Boot and arm
|
|
|
|
Create the U-Boot scipts at image buildtime and remove the binary
script representation from version control.
|
|
iMX6: Always do use a secondary argument for the 'mmc dev' command.
T30: Do rootfs update of eMMC in chunks as well. Requires pre-splitting
file in Ethernet case. As an advantage allows flashing rootfs bigger
than RAM plus avoids timeouts during flashing. Added wrapper flash_mmc
script as well for backwards compatibility reason.
While at it cleanup various whitespaces throughout the update scripts.
|
|
- rename flash_mmc script to flash_blk and adapt that it can also be used
from a USB drive.
- Add flash_mmc “proxy” script for the upgrade case
|
|
Also deploy an additional device tree supporting a Colibri iMX6 with
Analog Camera Module.
|
|
|
|
Unifing with imx6 update script pulled in a test that a device-tree
must be deployed. Apalis/Colibri T30 use a not device-tree enabled
Kernel. So remove the check that the device-tree actually exists.
|
|
The migration to BSP V2.4 is somewhat more complex due to ECC level
change. The scripts detect whether we are on a new or old module by
checking the NAND driver type (which probably reflects best which
ECC level has been used). In case the NAND driver has the old name
(fsl_nfc), then the script assumes that migration is required:
The migration script runs in two stages:
- The first part updates U-Boot, erases the root file system and
loads the config block into memory (upgrade_part1).
- After reboot, the second stage runs with the new U-Boot version.
The script update the boot configuration block (BCB), restores the
config block from memory and U-Boot is written again using the
new ECC level. Then, the script erases the UBI partition (which
contains the root filesystem) and recreates it using the new ECC
level too (upgrade_part2).
Since we keep the config block temporarly in RAM, the whole update
process should not be interrupted! In case it did, the new U-Boot
version is able to restore the configuration block from command
line.
In contrast to older update scripts, the standard upgrade procedure
does not erase the whole UBI partition anymore. This makes sure that
wear-leveling information can be preserved.
|
|
Make the option -o mandatary, used to default to /media/KERNEL
While at it synchronize the files again.
|
|
Update to latest version (SVN r671) of burnflash.
|
|
Update to latest version (r663) of Vybrid flash scripts.
|
