我这里试了一些ptp参数还是不行，这个问题似乎是系统本身配置的问题
目前我是将系统时钟源(/sys/devices/system/clocksource/clocksource0/current_clocksource)设置为usec临时规避，但是这个精度较差
系统默认的arch_sys_counter(硬件时钟源，同时JP512也是使用这个)和tsc(也是硬件时钟源)都会出现无规律的系统时间跳变，请问下是否可以从这个方向排查一下呢？
we are still checking this internally.
Hi,
Could you disable nvpps and try again?
To disable it, go to the device tree and search nvpps, disable the status there.
Thank you WayneWWW.
Significantly better than it was.
About every 12 tics phc offset jumps to around 120 then stays beneath 100 until it jumps again. I’ll try some different ptp4l and phc2sys command lines later.
cat disable-nvpps.dts
/dts-v1/;
/plugin/;
/ {
compatible = "nvidia,tegra234";
fragment@0 {
target-path = "/bus@0/nvpps";
__overlay__ {
status = "disabled";
};
};
};
dtc -I dts -O dtb -o disable-nvpps.dtbo disable-nvpps.dts
sudo cp ./disable-nvpps.dtbo /boot/
add these 2 lines after APPEND line in:
sudo nano /boot/extlinux/extlinux.conf
FDT /boot/dtb/kernel_tegra234-p3737-0000+p3701-0000-nv.dtb
OVERLAYS /boot/disable-nvpps.dtbo
sudo reboot
sudo ptp4l -f /etc/linuxptp/ptp4l.conf -i eno1 -m -2 -H
sudo phc2sys -s eno1 -c CLOCK_REALTIME -O 0 -m -w -S 0
Following has been running for 45 minutes and offset stays around to 20 to 60 and highest offset of between 100 and 120 for about 25 total seconds out of about 2700 seconds.
sudo ptp4l -f /etc/linuxptp/ptp4l.conf -i eno1 -m
sudo phc2sys -s eno1 -c CLOCK_REALTIME -w -m
cat /etc/linuxptp/ptp4l.conf
# /etc/linuxptp/ptp4l.conf
[global]
# Default Data Set
verbose 1
twoStepFlag 1
priority1 128
priority2 128
domainNumber 0
clockClass 6
clockAccuracy 0xFE
offsetScaledLogVariance 0xFFFF
utc_offset 37
maxStepsRemoved 255
# Port Data Set
logAnnounceInterval 1
logSyncInterval 0
announceReceiptTimeout 3
neighborPropDelayThresh 20000000
delayAsymmetry 0
# Run-time options
tx_timestamp_timeout 10
logging_level 6
kernel_leap 1
# Servo (PI)
clock_servo pi
pi_proportional_const 0.1
pi_integral_const 0.001
step_threshold 0.5
first_step_threshold 0.02
sanity_freq_limit 200000000
# Transport
network_transport L2
delay_mechanism E2E
time_stamping hardware
transportSpecific 0x0
# Timestamp processing
tsproc_mode filter
#delay_filter moving_average
#delay_filter_length 10
# Interface defaults
clock_type OC
egressLatency 0
ingressLatency 0
您好，WayneWWW，我这边disabled nvpps后，系统时间确实没有跳变的问题了
但是我们发现会出现另外的问题，Orin的4个XFI开始出现丢包问题，而且有一些丢包还很严重(使用ping就能看到明显的丢包)，如下图(192.168.5.16/64分别是两颗Orin模组的XFI0)：
然后我这边dtb的nvpps配置如下：
设备树文件如下：
dts.zip (39.5 KB)
请问下您这边有遇到吗？
No, we didn’t hit such issue.
Actually we don’t support 4 XFI on Orin AGX either. Only 1 XFI case is supported.