On 4/22/26 11:39 AM, Randolph Sapp wrote:
On Wed Apr 22, 2026 at 1:22 AM CDT, Yogesh Hegde wrote:
The am62lxx-evm board only has ttyS0 available. The k3 default
SERIAL_CONSOLES inherited includes ttyS2, which causes
systemd to wait indefinitely for a console that never appears,
significantly increasing boot time and degrading user experience.
Override SERIAL_CONSOLES to specify only ttyS0, allowing the boot
process to proceed normally without waiting for unavailable consoles.
Signed-off-by: Yogesh Hegde <[email protected]>
---
Boot time is increased from 33s to 2m18s.
Boot time is measured from power on to TI LVGL demo
running on HDMI Display.
Logs:
root@am62lxx-evm:~# systemd-analyze critical-chain ti-lvgl-demo.service
The time when unit became active or started is printed after the "@" character.
The time the unit took to start is printed after the "+" character.
ti-lvgl-demo.service @1min 32.588s
`-multi-user.target @1min 32.577s
`-getty.target @1min 32.575s
`[email protected] @13.002s
`-systemd-user-sessions.service @10.847s +660ms
`-network.target @10.582s
`-systemd-networkd.service @8.748s +1.799s
`-network-pre.target @8.735s
`-iptables.service @7.477s +920ms
`-basic.target @7.363s
`-dbus-broker.service @6.661s +466ms
`-dbus.socket @6.310s +434us
`-sysinit.target @6.263s
`-systemd-vconsole-setup.service @18.198s +391ms
`-systemd-journald.socket @2.781s +335us
`-system.slice @1.282s
`--.slice @1.282s
From the time analysis log, `getty.target` takes 1min 32 secs to bootup due to
[email protected] failed. This service fails because there is no
hardware associated with ttyS2.
---
meta-ti-bsp/conf/machine/am62lxx-evm.conf | 2 ++
1 file changed, 2 insertions(+)
diff --git a/meta-ti-bsp/conf/machine/am62lxx-evm.conf
b/meta-ti-bsp/conf/machine/am62lxx-evm.conf
index 677d6955..edfb4057 100644
--- a/meta-ti-bsp/conf/machine/am62lxx-evm.conf
+++ b/meta-ti-bsp/conf/machine/am62lxx-evm.conf
@@ -4,6 +4,8 @@
require conf/machine/include/am62lxx.inc
+SERIAL_CONSOLES = "115200;ttyS0"
+
IMAGE_FSTYPES += "ubifs ubi"
MKUBIFS_ARGS ?= "-F -m 4096 -e 258048 -c 12000"
Systemd's getty generator might still try to grab ttyS0 in other scenarios, but
this at least fixes the default behavior of trying to wait for the kernel serial
console on this board.
Out of curiosity, why did this change for this board? As I recall Andrew and
Nishanth had a debate about this a few years back and the result was to
standardize to ttyS2 for all k3.
AM62L isn't "K3" in a way, it doesn't have the R5 cores taking up a UART.
While we probably could have aliased the first MAIN domain UART to ttyS2
anyway, this feels like a good time to get back to the more standard
"default UART == ttyS0" that everyone else goes by.
Reviewed-by: Randolph Sapp <[email protected]>
-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.
View/Reply Online (#19871):
https://lists.yoctoproject.org/g/meta-ti/message/19871
Mute This Topic: https://lists.yoctoproject.org/mt/118950880/21656
Group Owner: [email protected]
Unsubscribe: https://lists.yoctoproject.org/g/meta-ti/unsub
[[email protected]]
-=-=-=-=-=-=-=-=-=-=-=-
