NXP LSDK linuxptp使用介绍

2022-09-05 本文已影响0人古斟布衣

NXP LSDK（Layerscape SDK）提供了linuxptp库，用于支持PTP/gPTP协议。本文基于LSDKv20.12，介绍了linuxptp协议栈的升级和使用方法。

1. linuxptp升级
2. linuxptp使用

1. linuxptp升级

根据packages/rfs/yocto-poky/meta/recipes-support/linuxptp/linuxptp_2.0.bb，找到linuxptp的下载网址为http://sourceforge.net/projects/linuxptp/files/v${PV}/linuxptp-${PV}.tgz，因此到相关网站上发现最新版本为http://sourceforge.net/projects/linuxptp/files/v3.1/linuxptp-3.1.1.tgz，因此进行如下修改以将linuxptp更新到最新版本：

将linuxptp_2.0.bb重命名为linuxptp_3.1.1.bb;

下载最新的linuxptp并解压，创建git仓库，根据packages/rfs/yocto-poky/meta/recipes-support/linuxptp/linuxptp目录下的2个patch修改文件并生成新的patch：

$ tar xvf linuxptp-3.1.1.tgz
$ cd linuxptp-3.1.1/
$ git init .
$ git add .
$ git commit -m "original version"
$ vim makefile                         # 修改Makefile
$ git commit                           # 第一次提交
$ vim makefile                         # 修改Makefile
$ git commit                           # 第二次提交
$ git log
commit 261bd281d79c4024a2413dac1d34bde1dc53ca2f (HEAD -> master)
Author: XXX <xxx@xxx.com>
Date:   Wed Mar 23 15:00:14 2022 +0800

      linuxptp: no incdefs using host headers

      Avoid using host headers via incdefs.sh shell script.

      Signed-off-by: XXX <xxx@xxx.com>

commit b18d6db45bafa6e82cd56208c07404908e71b428
Author: XXX <xxx@xxx.com>
Date:   Wed Mar 23 10:09:26 2022 +0800

      build: Allow CC and prefix to be overriden

      Signed-off-by: XXX <xxx@xxx.com>

commit 1e4cbdfb164b4bc1bfa9fbbd95e62faafb11f8f4
Author: XXX <xxx@xxx.com>
Date:   Wed Mar 23 09:54:52 2022 +0800

      original version
$ git format-patch HEAD^^
0001-build-Allow-CC-and-prefix-to-be-overriden.patch
0002-linuxptp-no-incdefs-using-host-headers.patch

用生成的patch替换packages/rfs/yocto-poky/meta/recipes-supportlinuxptp/linuxptp目录下原有的patch；

修改linuxptp_3.1.1.bb，更新下载连接、文件校验码和patch名称：

SRC_URI = "http://sourceforge.net/projects/linuxptp/files/v3.1linuxptp-${PV}.tgz \
       file://0001-build-Allow-CC-and-prefix-to-be-overridenpatch \
       file://0002-linuxptp-no-incdefs-using-host-headers.patch\
       "

SRC_URI[md5sum] = "3b79ab5e77c5b5cf06bc1c8350d405bb"
SRC_URI[sha256sum] ="94d6855f9b7f2d8e9b0ca6d384e3fae6226ce6fc012dbad02608bdef3be1c09"

gPTP配置文件在linuxptp目录：

flexbuild_lsdk2012$ find . -name gPTP.cf
./packages/rfs/yocto-poky/build/tmp/work/core2-64-poky-linux/linuxptp/2.0-r0/linuxptp-3.1.1/configs/gPTP.cfg

存在多个版本时，可以在configs/yocto/local_arm64_devel.conf文件中指定版本号：

@@ -37,6 +37,8 @@ DISTRO ?= "poky"
 PACKAGE_CLASSES ?= "package_deb"
+PREFERRED_VERSION_linuxptp = "3.1.1"
+
 IMAGE_FSTYPES = "cpio.gz ubi squashfs"

修改linxuptp包，打包文件系统时自动将gPTP.cfg拷贝到/usr/bin目录：

diff --git a/meta/recipes-extra/linuxptp/linuxptp/gPTP.cfg b/meta/recipes-extra/linuxptp/linuxptp/gPTP.cfg
new file mode 100644
...
diff --git a/meta/recipes-extra/linuxptp/linuxptp_3.1.1.bb b/meta/recipes-extra/linuxptp/linuxptp_3.1.1.bb
index c971926a53..2bb91a8a5f 100755
--- a/meta/recipes-extra/linuxptp/linuxptp_3.1.1.bb
+++ b/meta/recipes-extra/linuxptp/linuxptp_3.1.1.bb
@@ -5,7 +5,7 @@ LIC_FILES_CHKSUM = "file://COPYING;md5=b234ee4d69f5fce4486a80fdaf4a4263"
 SRC_URI = "http://sourceforge.net/projects/linuxptp/files/v3.1/linuxptp-${PV}.tgz \
            file://0001-build-Allow-CC-and-prefix-to-be-overriden.patch \
            file://0002-linuxptp-no-incdefs-using-host-headers.patch \
-           "
+           file://gPTP.cfg"

 SRC_URI[md5sum] = "3b79ab5e77c5b5cf06bc1c8350d405bb"
 SRC_URI[sha256sum] = "94d6855f9b7f2d8e9b0ca6d384e3fae6226ce6fc012dbad02608bdef3be1c0d9"
@@ -16,6 +16,7 @@ EXTRA_OEMAKE = "ARCH=${TARGET_ARCH} \
 do_install () {
     install -d ${D}/${bindir}
     install -p ${S}/ptp4l  ${D}/${bindir}
+    install -m0600 ${WORKDIR}/gPTP.cfg ${D}${bindir}/gPTP.cfg
     install -p ${S}/pmc  ${D}/${bindir}
     install -p ${S}/phc2sys  ${D}/${bindir}
     install -p ${S}/hwstamp_ctl  ${D}/${bindir}

编译文件系统，具体细节参加NXP LSDK简明实用手册的正式文件系统编译和文件系统镜像创建两章：
```
flexbuild_lsdk2012$ source setup.env
flexbuild_lsdk2012$ flex-builder -m lx2160ardb_rev2 -a arm64 -i mkrfs -r yocto:devel
```
更新文件系统，并下载到板卡上：
```
root@DevelLinux:~# ptp4l -v
3.1.1
```

2. linuxptp使用

将2台设备的待测试网口对接；

在2台设备上分别使用如下命令进行PTP测试：

root@DevelLinux:~# ifconfig eth0 192.168.200.101/24 up && ptp4l -i eth0 -m      # 设备1
ptp4l[18.866]: selected /dev/ptp0 as PTP clock
ptp4l[18.916]: port 1: INITIALIZING to LISTENING on INIT_COMPLETE
ptp4l[18.917]: port 0: INITIALIZING to LISTENING on INIT_COMPLETE
ptp4l[18.917]: port 1: link down
ptp4l[18.917]: port 1: LISTENING to FAULTY on FAULT_DETECTED (FT_UNSPECIFIED)
ptp4l[18.926]: selected local clock 0629fa.fffe.fd2d1b as best master
ptp4l[18.926]: port 1: assuming the grand master role
[   21.067516] dpaa2_mac dpmac.7 mac7: Link is Up - 1Gbps/Full - flow control off
[   21.080889] fsl_dpaa2_eth dpni.4 eth0: Link Event: state up
[   21.086775] IPv6: ADDRCONF(NETDEV_CHANGE): eth0: link becomes ready
ptp4l[20.918]: port 1: link up
ptp4l[20.978]: port 1: FAULTY to LISTENING on INIT_COMPLETE
ptp4l[27.897]: port 1: LISTENING to MASTER on ANNOUNCE_RECEIPT_TIMEOUT_EXPIRES
ptp4l[27.897]: selected local clock 0629fa.fffe.fd2d1b as best master
ptp4l[27.897]: port 1: assuming the grand master role
ptp4l[28.113]: port 1: new foreign master 2e19fe.fffe.2e8922-1
[   60.198448] ffs_data_put(): freeing
...
root@DevelLinux:~# ifconfig eth0 192.168.200.102/24 up && ptp4l -i eth0 -m    # 设备2
ptp4l[15.750]: selected /dev/ptp0 as PTP clock
ptp4l[15.799]: port 1: INITIALIZING to LISTENING on INIT_COMPLETE
ptp4l[15.800]: port 0: INITIALIZING to LISTENING on INIT_COMPLETE
ptp4l[15.800]: port 1: link down
ptp4l[15.800]: port 1: LISTENING to FAULTY on FAULT_DETECTED (FT_UNSPECIFIED)
ptp4l[15.810]: selected local clock 2e19fe.fffe.2e8922 as best master
ptp4l[15.810]: port 1: assuming the grand master role
[   17.936870] dpaa2_mac dpmac.7 mac7: Link is Up - 1Gbps/Full - flow control off
[   17.950549] fsl_dpaa2_eth dpni.4 eth0: Link Event: state up
[   17.956292] IPv6: ADDRCONF(NETDEV_CHANGE): eth0: link becomes ready
ptp4l[17.786]: port 1: link up
ptp4l[17.848]: port 1: FAULTY to LISTENING on INIT_COMPLETE

ptp4l[25.415]: port 1: LISTENING to MASTER on ANNOUNCE_RECEIPT_TIMEOUT_EXPIRES
ptp4l[25.415]: selected local clock 2e19fe.fffe.2e8922 as best master
ptp4l[25.415]: port 1: assuming the grand master role
[   33.296169] dpaa2_mac dpmac.7 mac7: Link is Down
[   33.302513] fsl_dpaa2_eth dpni.4 eth0: Link Event: state down
ptp4l[33.416]: timed out while polling for tx timestamp
ptp4l[33.417]: increasing tx_timestamp_timeout may correct this issue, but it is likely caused by a driver bug
ptp4l[33.417]: port 1: send sync failed
ptp4l[33.417]: port 1: MASTER to FAULTY on FAULT_DETECTED (FT_UNSPECIFIED)
[   34.320171] dpaa2_mac dpmac.7 mac7: Link is Up - 1Gbps/Full - flow control off
ptp4l[34.144]: port 1: link down
ptp4l[34.151]: selected local clock 2e19fe.fffe.2e8922 as best master
ptp4l[34.151]: port 1: assuming the grand master role
[   34.338790] fsl_dpaa2_eth dpni.4 eth0: Link Event: state up
ptp4l[34.171]: port 1: link up
ptp4l[34.237]: port 1: FAULTY to LISTENING on INIT_COMPLETE
ptp4l[41.312]: port 1: new foreign master 0629fa.fffe.fd2d1b-1
ptp4l[41.525]: port 1: LISTENING to MASTER on ANNOUNCE_RECEIPT_TIMEOUT_EXPIRES
ptp4l[41.525]: selected local clock 2e19fe.fffe.2e8922 as best master
ptp4l[41.526]: port 1: assuming the grand master role
ptp4l[45.312]: selected best master clock 0629fa.fffe.fd2d1b
ptp4l[45.312]: port 1: MASTER to UNCALIBRATED on RS_SLAVE
ptp4l[48.311]: master offset 13413912505 s0 freq      +0 path delay      7315
ptp4l[49.311]: master offset 13413909795 s1 freq   -2710 path delay      7315
ptp4l[50.311]: master offset      -1344 s2 freq   -4054 path delay      7315
ptp4l[50.312]: port 1: UNCALIBRATED to SLAVE on MASTER_CLOCK_SELECTED
ptp4l[51.312]: master offset        562 s2 freq   -2551 path delay      6751
ptp4l[52.312]: master offset        256 s2 freq   -2688 path delay      6791
ptp4l[53.312]: master offset         80 s2 freq   -2788 path delay      6781
ptp4l[54.312]: master offset         84 s2 freq   -2760 path delay      6761
...

注意，此时使用默认配置，即packages/rfs/yocto-poky/build/tmp/work/core2-64-poky-linux/linuxptp/2.0-r0/linuxptp-3.1.1/configs/default.cfg。

根据LSDK手册，删除gPTP.cfg中的neighborPropDelayThresh 800配置，并将其下载到板子的/home/root目录，然后在2台设备上分别使用如下命令进行gPTP测试：

root@DevelLinux:~# ifconfig eth0 192.168.200.101/24 up && ptp4l -i eth0 -f gPTP.cfg-m &    # 设备1
ptp4l[22.408]: selected /dev/ptp0 as PTP clock
ptp4l[22.481]: port 1: INITIALIZING to LISTENING on INIT_COMPLETE
ptp4l[22.481]: port 0: INITIALIZING to LISTENING on INIT_COMPLETE
ptp4l[22.481]: port 1: link down
ptp4l[22.481]: port 1: LISTENING to FAULTY on FAULT_DETECTED (FT_UNSPECIFIED)
ptp4l[22.536]: selected local clock 42ac62.fffe.6db1ab as best master
ptp4l[22.536]: port 1: assuming the grand master role
[   24.587754] dpaa2_mac dpmac.7 mac7: Link is Up - 1Gbps/Full - flow control off
[   24.601130] fsl_dpaa2_eth dpni.4 eth0: Link Event: state up
[   24.606900] IPv6: ADDRCONF(NETDEV_CHANGE): eth0: link becomes ready
ptp4l[24.437]: port 1: link up
ptp4l[24.518]: port 1: FAULTY to LISTENING on INIT_COMPLETE
ptp4l[27.805]: port 1: LISTENING to MASTER on ANNOUNCE_RECEIPT_TIMEOUT_EXPIRES
ptp4l[27.806]: selected local clock 42ac62.fffe.6db1ab as best master
ptp4l[27.806]: port 1: assuming the grand master role
[   35.851758] dpaa2_mac dpmac.7 mac7: Link is Down
[   35.858121] fsl_dpaa2_eth dpni.4 eth0: Link Event: state down
ptp4l[36.519]: timed out while polling for tx timestamp
ptp4l[36.520]: increasing tx_timestamp_timeout may correct this issue, but it is likely caused by a driver bug
ptp4l[36.520]: port 1: send peer delay request failed
ptp4l[36.520]: port 1: MASTER to FAULTY on FAULT_DETECTED (FT_UNSPECIFIED)
[   36.875755] dpaa2_mac dpmac.7 mac7: Link is Up - 1Gbps/Full - flow control off
ptp4l[36.700]: port 1: link down
ptp4l[36.707]: selected local clock 42ac62.fffe.6db1ab as best master
ptp4l[36.707]: port 1: assuming the grand[   36.893879] fsl_dpaa2_eth dpni.4 eth0: Link Event: state up
master role
ptp4l[36.726]: port 1: link up
ptp4l[36.803]: port 1: FAULTY to LISTENING on INIT_COMPLETE
ptp4l[40.092]: port 1: LISTENING to MASTER on ANNOUNCE_RECEIPT_TIMEOUT_EXPIRES
ptp4l[40.092]: selected local clock 42ac62.fffe.6db1ab as best master
ptp4l[40.092]: port 1: assuming the grand master role
ptp4l[41.426]: port 1: new foreign master a6ef06.fffe.60084e-1
...
root@DevelLinux:~# ifconfig eth0 192.168.200.102/24 up && ptp4l -i eth0 -f gPTP.cfg -m &# 设备2
ptp4l[17.390]: selected /dev/ptp0 as PTP clock
ptp4l[17.453]: port 1: INITIALIZING to LISTENING on INIT_COMPLETE
ptp4l[17.453]: port 0: INITIALIZING to LISTENING on INIT_COMPLETE
ptp4l[17.453]: port 1: link down
ptp4l[17.453]: port 1: LISTENING to FAULTY on FAULT_DETECTED (FT_UNSPECIFIED)
ptp4l[17.504]: selected local clock a6ef06.fffe.60084e as best master
ptp4l[17.504]: port 1: assuming the grand master role
[   19.563863] dpaa2_mac dpmac.7 mac7: Link is Up - 1Gbps/Full - flow control off
[   19.577303] fsl_dpaa2_eth dpni.4 eth0: Link Event: state up
[   19.582908] IPv6: ADDRCONF(NETDEV_CHANGE): eth0: link becomes ready
ptp4l[19.413]: port 1: link up
ptp4l[19.494]: port 1: FAULTY to LISTENING on INIT_COMPLETE
ptp4l[22.830]: port 1: new foreign master 42ac62.fffe.6db1ab-1
ptp4l[23.161]: port 1: LISTENING to MASTER on ANNOUNCE_RECEIPT_TIMEOUT_EXPIRES
ptp4l[23.161]: selected local clock a6ef06.fffe.60084e as best master
ptp4l[23.161]: port 1: assuming the grand master role
ptp4l[24.830]: selected best master clock 42ac62.fffe.6db1ab
ptp4l[24.830]: port 1: MASTER to UNCALIBRATED on RS_SLAVE
ptp4l[25.830]: port 1: UNCALIBRATED to SLAVE on MASTER_CLOCK_SELECTED
ptp4l[26.455]: rms 9131880424 max 18263760993 freq   +791 +/- 648 delay  6733 +/-   0
ptp4l[27.456]: rms  166 max  329 freq  +1475 +/- 203 delay  6730 +/-   0
ptp4l[28.456]: rms  198 max  220 freq  +1936 +/-  67 delay  6729 +/-   0
ptp4l[29.457]: rms  170 max  210 freq  +2093 +/-  43 delay  6729 +/-   0
ptp4l[30.457]: rms   84 max  110 freq  +2103 +/-  21 delay  6729 +/-   0
...

注意，此时使用gPTP.cfg覆盖默认配置中的部分选项。

每台设备上可以同时启动多个gPTP程序，但是不能存在多个gPTP master，否则会自动切换；

root@DevelLinux:~# ifconfig eth0 192.168.200.101/24 up && ptp4l -i eth0 -f gPTP.cfg-m &    # 设备1
root@DevelLinux:~# ifconfig eth0 192.168.200.102/24 up && ptp4l -i eth0 -f gPTP.cfg-m &    # 设备2
root@DevelLinux:~# ifconfig eth2 192.168.201.101/24 up && ptp4l -i eth2 -f gPTP.cfg-m &    # 设备1
root@DevelLinux:~# ifconfig eth1 192.168.201.102/24 up && ptp4l -i eth1 -f gPTP.cfg-m &    # 设备2

根据NXP建议，在gPTP.cfg中增加如下配置项，并为从节点启动命令增加-s选项以限制模式为slave，未发现明显改进：
```
neighborPropDelayThresh 20000
summary_interval        -3
```

根据NXP建议，CPU频率是影响测试结果的重要因素，因此查看CPU支持频率和当前频率：

root@DevelLinux:~# cat /sys/devices/system/cpu/cpu0/cpufreqscaling_available_frequencies
1800000 900000
root@DevelLinux:~# cat /sys/devices/system/cpu/cpu0/cpufreqscaling_cur_freq
1800000

当前频率已经是当前RCW可以支持的最高频率，否则可以使用如下命令调整：

root@DevelLinux:~# echo 1800000 > /sys/devices/system/cpu/cpu0cpufreq/scaling_setspeed
1800000

NXP LSDK linuxptp使用介绍

1. linuxptp升级

2. linuxptp使用

猜你喜欢

热点阅读