On 2/12/2026 8:52 AM, Jon Cormier wrote:
On Wed, Feb 11, 2026 at 5:05 PM Denys Dmytriyenko via
lists.yoctoproject.org <[email protected]> wrote:
On Wed, Feb 11, 2026 at 03:57:28PM -0600, Ryan Eatmon via
lists.yoctoproject.org wrote:
On 2/11/2026 2:25 PM, Denys Dmytriyenko wrote:
On Wed, Feb 11, 2026 at 02:04:14PM -0600, Ryan Eatmon via
lists.yoctoproject.org wrote:
Clean up the logic of the new ti-core-initramfs to reduce the impact on
downstream layers.
- Flip the logic in the machine .conf files to track the list of
required kernel modules for that platform, but if the variable is set
to "" then the ti-core-initramfs will not be created.
- Redo the logic of when we require creation/usage of the initramfs to
just platforms that need it (ie that set TI_CORE_INITRAMFS_KERNEL_MODULES).
- Add a big switch to disable the initramfs entirely. In that case, the
user is on their own to make sure that the kernel has everything it
needs to boot either via config fragments to turn on the needed
modules, or by using the TI_CORE_INITRAMFS_KERNEL_MODULES
variable to populate their own initramfs.
In our discussion we agreed to use TI_CORE_INITRAMFS_KERNEL_MODULES to
determine whether to enable initramfs or not. While I don't mind adding
a separate global TI_CORE_INITRAMFS_ENABLED switch, but I now wonder if
two checks are redundant and whether it makes sense to create initramfs
even when the list of modules is empty, if it's enabled?
I thought about this and went this way for a reason. If you want to
turn off the initramfs, but you still want the list of what modules
are required say for your own initramfs, then the list is gone.
Because the list was the switch.
By splitting them into two lists, we gain the fact that we don't
force the initramfs on platforms that don't need it, and you can
still turn it all off if you want but have all of the infrastructure
in place to use it in your own layer later.
I think you missed my point - I'm not against the "enable" flag, but since we
now have it explicitly, does it make sense to also build the initramfs with an
empty modules list when the "enable" flag is set to "1"?
Being able to create an initramfs without modules makes sense to me.
We've had to use initramfs before for mounting filesystems in a
particular order for example. And didn't really need extra modules at
the time. Just a thought.
I am going to send a second patch that adds support for adding
additional packages to the ti-core-initramfs that also turns it on.
That way a distro can chose to always turn it on and add in that kind of
change.
Also see below for another comment.
+++ b/meta-ti-bsp/conf/machine/include/ti-core-initramfs.inc
@@ -0,0 +1,29 @@
+#------------------------------------------------------------------------------
+#
+# To turn off the ti-core-initramfs.cpio creation just set:
+#
+# TI_CORE_INITRAMFS_ENABLED = "0"
+#
+#------------------------------------------------------------------------------
+TI_CORE_INITRAMFS_ENABLED ?= "0"
Did you mean to weakly set it to "1" here? :)
Doh. I did. That was from testing. I'll send a v2. Good catch.
--
Ryan Eatmon [email protected]
-----------------------------------------
Texas Instruments, Inc. - LCPD - MGTS
-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.
View/Reply Online (#19502):
https://lists.yoctoproject.org/g/meta-ti/message/19502
Mute This Topic: https://lists.yoctoproject.org/mt/117763070/21656
Group Owner: [email protected]
Unsubscribe: https://lists.yoctoproject.org/g/meta-ti/unsub
[[email protected]]
-=-=-=-=-=-=-=-=-=-=-=-
