find / -name pinmux-pins
cat pinmux-pins
find / -name clk_summary
cat clk_summary
find / -name regulator_summary
cat regulator_summary
tar cvf dts.tar dts,使用的dts目录打包发过来,指明使用哪个dts
find / -name pinmux-pins cat pinmux-pins 显示的信息一起打包发过来
cat /sys/kernel/debug/d/clk/clk_summary 显示的信息一起打包发过来
开两个终端,一个查看Kconfig,一个在menuconfig上搜索(所有的Kconfig子块可以搜索到),可以很快确定目前的模块驱动
一种是我操作ok了你硬件不行产生偏差(主要是25CLK频偏),一种是我操作ok你没反应就像石头一样没有启动(phy没有启动,rest引脚或者vcc或者有些线没有接),一种是我操作有问题比如我被复用,
一种是我操作ok你也ok但是有干扰(走线干扰,delay或者电压比如概率性获取不到ip或者丢包或者断线)
确认mac clk tx clk rx clk mdc clk,再确认复用,gmac奇怪问题比较多是电源的问题,特别是CPU LOGIC电压的问题或者IO驱动强度,配了CRU clk却拿不到,
查看clk_summary没有一个是使能的
PHY的问题主要就是25M晶振,没有贴没有clk给出,txclk是由phy clk给gmac,然后gmac再给出去tx clk,有管脚复用设置phy的时候可能RX clk也不对
总之是gmac操作phy的寄存器(rx clk是操作phy寄存器给出的,mdc clk是gmac给出的【源头CRU 内部clk cpu 25M晶振 或者phy 25M晶振】),
clk数量多了(速率快干扰错),包括sdio phy uart降频 ,减少clk数量,降低看是否排除干扰
0、确认一下phy上晶振25M晶振频偏,网线插路由能否获取到IP,如果不能检查一下硬件,如果能用iperf测一下吞吐
1、对于有些差的网络,尤其公司网络,同一个DHCP服务器下面主机过多,有时候会出现,DHCP获取不到IP地址的现象。
对于这种问题,处理办法是,把DHCP的超时时间增大。目前的dhcp超时时间是30s,补丁把超时时间增大到90s。 插拔网卡出现概率性性获取不了ip
89785 参考 把超时时间增大到90s.patch
2、也测量一下cpu logic 电压 看一下dts上面的PMU的配置 正常是1.05v左右
regulators {
compatible = "simple-bus";
#address-cells = <1>;
#size-cells = <0>;
vdd_logic: regulator@0 {
compatible = "regulator-fixed";
regulator-name = "vdd_logic";
regulator-min-microvolt = <1050000>;
regulator-max-microvolt = <1050000>;
regulator-always-on;
};
};
PMU的电压输出在sys/kernel/debug/regulator/
这里可以手动调试一下
cat /sys/kernel/debug/regulator/vdd_logic/*
cat /sys/kernel/debug/regulator/vdd_log/*
echo 1050000 > /sys/kernel/debug/regulator/vdd_log/voltage
logic电压是根据ddr的频率调整的,当前的ddr频率是多少?切换一下ddr频率看是否有变化
root@rk3399_stbvr:/sys/kernel/debug/clk # cat clk_summary |grep ddr
clk_pvtm_ddr 0 0 24000000 0 0
pclk_ddr 1 1 200000000 0 0
pclk_ddr_sgrf 0 0 200000000 0 0
pclk_ddr_mon 0 0 200000000 0 0
clk_ddrc_gpll_src 0 0 800000000 0 0
clk_ddrc_dpll_src 1 1 792000000 0 0
sclk_ddrc 1 1 792000000 0 0
clk_ddrc_bpll_src 0 0 408000000 0 0
clk_ddrc_lpll_src 0 0 408000000 0 0
ddr 800M,vdd_log为0.9V,arch/arm64/boot/dts/rockchip/rk3399-opp.dtsi里面dmc_opp_table,改800M对应的电压opp-microvolt
opp-800000000 {
opp-hz = /bits/ 64 <800000000>;
opp-microvolt = <900000>;
TX上不去是DDR不够,定频到800M,TX就上去了
驱动强度的还是改一下
cat clk_summary |grep ddr
sclk_ddrc就是DDR的频率
cat /sys/kernel/debug/opp/opp_summary
cd /sys/class/devfreq/dmc/
echo userspace > governor
echo 800000000 > userspace/set_freq
ddr频率对以太网有影响,考虑到休眠变频的影响或者本身上行低
&dmc {
status = "okay";
center-supply = <&vdd_center>;
upthreshold = <40>;
downdifferential = <20>;
system-status-freq = <
/*system status freq(KHz)*/
SYS_STATUS_NORMAL 800000
SYS_STATUS_REBOOT 800000
SYS_STATUS_SUSPEND 800000
SYS_STATUS_VIDEO_1080P 800000
SYS_STATUS_VIDEO_4K 800000
SYS_STATUS_VIDEO_4K_10B 800000
SYS_STATUS_PERFORMANCE 800000
SYS_STATUS_BOOST 800000
SYS_STATUS_DUALVIEW 800000
SYS_STATUS_ISP 800000
>;
3、上行或者下行带宽比较低,调试相关delay和硬件没有用,和TX RX CLK息息相关
这块板子走线有那么长tx_clk 13mA驱动强度不够,软件调整为19mA
--- a/arch/arm64/boot/dts/rockchip/rk3399.dtsi
+++ b/arch/arm64/boot/dts/rockchip/rk3399.dtsi
+ pcfg_pull_none_19ma: pcfg-pull-none-19ma {
+ bias-disable;
+ drive-strength = <19>;
+ };
rgmii_pins: rgmii-pins {
rockchip,pins =
/* mac_txclk */
- <3 17 RK_FUNC_1 &pcfg_pull_none_13ma>,
+ <3 17 RK_FUNC_1 &pcfg_pull_none_19ma>,
一般设置tx0~3 tx clk tx en 6个脚到19ma
4、设置静态
ifconfig eth0 up
ifconfig eth0 192.168.1.100 broadcast 192.168.1.255 netmask 255.255.255.0 up
单网卡添加多个IP地址
ifconfig eth0:0 192.168.1.100 netmask 255.255.255.0 up
ifconfig eth0:1 192.168.2.100 netmask 255.255.255.0 up
若转换不了域名 主要是域名问题
echo "nameserver 8.8.8.8" > /etc/resolv.conf
或者
ndc resolver setnetdns eth0 "" 8.8.8.8
用ip route
ip addr add 192.168.1.100/24 dev eth0
ip link set dev eth0 up
echo "nameserver 8.8.8.8" > /etc/resolv.conf或者ndc resolver setnetdns eth0 "" 8.8.8.8 8.8.4.4
ip route add default via 192.168.1.1 dev eth0
网关 网关 网关
和IP 地址卵关系都没有
route -n
route del default gw 192.168.1.1 dev eth0
route add default gw 192.168.1.1 dev eth0 metric 99
5、cat /sys/devices/platform/fe300000.ethernet/net/eth0/carrier
拔插网线的前后输入这个命令,看下底层状态是否正确;
平台不同节点路径可能不一样,在sys下面输入find -name "carrier" 没插网线的时候是0 ,插网线的时候是1
6、rtl8152有自带驱动的kernel/drivers/net/usb,无法拿到IP 地址,可能是没有MAC 地址的问题,
可使用ftp 上的补丁《rtl8152解决MAC 地址为空补丁》
7、以太网抓包
tcpdump -i eth0 -s 0 -w /data/snf.pcap
tcpdump -i wlan0 -s 0 -w /data/wlan.pcap
,抓p2p,关掉wifi,抓一下正常和异常的log
tcpdump -i any -s 0 -w /data/tcpdump_log.pcap
8、过滤
EthernetNetworkFactory /frameworks/opt/net/ethernet/java/com/android/server/ethernet/
ConnectivityService
https://blog.csdn.net/u013686019/article/details/51447129
或者网络优先级更改
9、PHY ID 00000000 就要确认一下mdc clk的2.5M以下(可以打上mdc2.5M patch)
对照原理图,先确定IOMUX关系是否正确
busybox find /d/pinctrl -name pinmux-pins
cat ./sys/kernel/debug/pinctrl/pinctrl/pinmux-pins
[ 0.668216] eth%d: PHY ID 00000000 at 1 IRQ POLL (stmmac-0:01)
PHY ID不正确,示波器量测MDC是2.5M 及2.5M以下,使用patch使MDC clk降低
10、某些交换机上面之后出现不能正常获取到IP地址
目前出现一些板卡,接在某些交换机上面之后出现不能正常获取到IP地址,频繁重复获取,有些交换机或者是路由器又是正常,可能跟EEEE功能有关,
打上关闭EEE功能的patch
11、cat /d/clk/clk_summary
对照原理图,先确定IOMUX关系是否正确,
busybox find /d/pinctrl -name pinmux-pins
cat xxx/xxx/pinmux-pins
12、RMII_CLK_CTL配置是低,PHY就会有50M输出,如果是拉高就是我们给PHY时钟
13、3399经常MAC MDC引脚会被复用,测的频率还是TX CLK 125M, RX CLK 25M(异常,导致phy设置异常),MAC CLK 125M
GPIO3B[0]
gslx680: gslx680@40 {
touch-gpio = <&gpio3 RK_PB0 IRQ_TYPE_EDGE_RISING>;
14、
ping -s 1024 xx.xx.xx.xx
ping -s 指定 ping包的大小 从 1k 往上加 ,确认ping 包方式是否 在ping的数据包增大到一定程度 也会出问题
15、双以太网目前补丁就是eth0用于访问外网,eth1或usb0只能访问局域网
给客户建议,设置里面加个切换按钮,设置用哪个网口上网,保存到属性里,然后
EthernetNetworkFactory.java和EthernetNetworkFactoryExt.java两个文件根据属性来切换接口
EthernetNetworkFactory这个文件中的可以访问外网,自行参考和开发需求
16、识别网卡是MDC CLK和MDIO两条线,识别不了MDC就没有MDC CLK了,no phy found和硬件关系较大
17、phy_register查看
kernel 3.10
find ./ -name phy_reg
cat phy_reg
kernel 4.4
find ./ -name phy_registers
cat phy_registers
18、不能自动获取IP的主板,手动设置静态IP能正常上网
8201F是百兆的PHY,不走delay节点,看PCB布线RMII线也比较短
示波器看频偏值不准确,要用频率计测试
指定IP后IPERF测吞吐量能到多少?如吞吐量带宽能达到94M左右,硬件测一下RJ45眼图,看电平是否达标,如眼图测试没问题。可能是DHCP服务有问题,另外也可以用“busybox udhcpc -i eth0”或“dhcptool eth0 ”看能否自动获取到IP
19、获取IP问题也需要多环境测试,也有是本身环境的问题
可能每个工位都强制设了静态IP的原因 所以才会有时能获取,有时候获取不了是因为IP冲突,单独试了WIFI路由器出来的每个主板都是可以
20、概率性获取不到ip(表现为dhcp timeout)
第一种是开关机,解决:确认reset引脚先拉低延时再拉高延时
第二种是不断插拔,解决:确认一下logic电压,tx rx delay修改吞吐测试,io电流加大,dhcp延时加一下,或者是变压器硬件的原因
第三种是休眠唤醒之后或者长时间概率性,解决:disable eee patch可能是phy的原因节能标准不一致,确认logic电压,驱动唤醒函数,还不行ifconfig eth0 down/up 补丁
21、唤醒后获取不了ip
diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
index a534672d0955..e60d34e294ad 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
@@ -3142,6 +3142,10 @@ int stmmac_resume(struct device *dev)
if (priv->phydev)
phy_start(priv->phydev);
+ stmmac_release(ndev);
+ msleep(100);
+ stmmac_open(ndev);
+
return 0;
}
EXPORT_SYMBOL_GPL(stmmac_resume);
22、用stmicro驱动时修改最大速率限制
--- a/arch/arm/boot/dts/rk3288-evb.dtsi
+++ b/arch/arm/boot/dts/rk3288-evb.dtsi
@@ -229,7 +229,7 @@
pinctrl-0 = <&rgmii_pins>;
tx_delay = <0x30>;
rx_delay = <0x10>;
- max-speed = <100>;
+ max-speed = <1000>;
status = "okay";
};
23、uboot以太网
u-boot/drivers/net
designware.o gmac_rockchip.o
CONFIG_DM_ETH=y
CONFIG_ETH_DESIGNWARE=y
CONFIG_GMAC_ROCKCHIP=y
CONFIG_DM_REGULATOR=y
CONFIG_DM_GPIO=y
24、关于休眠唤醒
休眠一下子就被唤醒了,断开负载发现和还是会出这种问题
亲测,蓝牙电源关掉,就可以了,蓝牙电源打开,PMEB唤醒引脚接地也可以进入休眠,说明不是蓝牙唤醒的
亲测,有些硬件蓝牙电源关掉也没有用,这就是个大坑,就是休眠模块的问题,IOMUX确认过的,IO domain在休眠也打开关闭都已经测试
波形也抓了没有问题,IO电流还是系统供电问题可能有关,就是休眠模块的坑
所以其它模块(蓝牙rfkill ethernet),方法都是去申请一个脚,然后去添加唤醒函数
disable_irq(irq->irq);
ret = enable_irq_wake(irq->irq);
然后在休眠的时候enable_irq
唤不醒先断开负载,对比一下IO DOMIN和 VDDIO是否匹配,不行也休眠模块的坑
25、推流卡顿
ethtool -k eth0 查看一下
关掉TX的时候的checksum,关闭硬件校验
ethtool -K eth0 tx-checksum-ipv4 off
ethtool -K eth0 tx-checksum-ipv6 off
关闭tso
ethtool ‐K eth0 tx‐tcp‐segmentation off
ethtool ‐K eth0 tx‐tcp6‐segmentation off
--- a/drivers/net/ethernet/rockchip/gmac/stmmac_main.c
+++ b/drivers/net/ethernet/rockchip/gmac/stmmac_main.c
@@ -3151,7 +3151,7 @@ struct stmmac_priv *stmmac_dvr_probe(struct device *device,
ndev->netdev_ops = &stmmac_netdev_ops;
- ndev->hw_features = NETIF_F_SG | NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM |
+ ndev->hw_features = NETIF_F_SG | NETIF_F_IPV6_CSUM |
NETIF_F_RXCSUM;
ndev->features |= ndev->hw_features | NETIF_F_HIGHDMA;
ndev->watchdog_timeo = msecs_to_jiffies(watchdog);
26、delay确认
--- a/drivers/net/ethernet/stmicro/stmmac/dwmac-rk.c
+++ b/drivers/net/ethernet/stmicro/stmmac/dwmac-rk.c
@@ -418,6 +419,8 @@ static void rk3288_set_to_rgmii(struct rk_priv_data *bsp_priv,
return;
}
+ printk("%s tx_delay=%x rx_delay=%x\n",__func__,tx_delay,rx_delay);
+
regmap_write(bsp_priv->grf, RK3288_GRF_SOC_CON1,
RK3288_GMAC_PHY_INTF_SEL_RGMII |
RK3288_GMAC_RMII_MODE_CLR);
@@ -1570,6 +1573,15 @@ static int rk_gmac_resume(struct device *dev)
27、以太网MAC地址ieee申请
一般卖到国外才需要申请
http://blog.sina.com.cn/s/blog_7d02693c0102v77o.html
28、以太网probe
4.4
stmmac_pltfr_probe
3.10
rk_gmac_probe
29、endpoint
iperf -c 192.168.32.252 -i 1 -t 99 -d
-d 同时进行双向传输测试,双向能同时到800M以上, 总吞吐量是两条流相加,有可能跟工具有关系。Marvell1512 导致16%CPU挂满,很少见
用endpoint iocharoit 测试
30、interrupts,CPU资源不够导致iperf测试吞吐低
cat /proc/interrupts
CPU0 CPU1 CPU2 CPU3 CPU4 CPU5
24: 84760 0 0 0 0 0 GICv3 44 Level eth0
echo 4 > /proc/irq/24/smp_affinity_list(换挂到CPU4)
24: 84760 0 0 0 393058 0 GICv3 44 Level eth0
diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
index abde6c5..791c32c 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
@@ -2833,8 +2833,10 @@ int stmmac_dvr_probe(struct device *device,
struct stmmac_resources *res)
{
int ret = 0;
+ int cpu_id =5;
struct net_device *ndev = NULL;
struct stmmac_priv *priv;
+ struct cpumask cpumask;
ndev = alloc_etherdev(sizeof(struct stmmac_priv));
if (!ndev)
@@ -2855,7 +2857,11 @@ int stmmac_dvr_probe(struct device *device,
priv->dev->irq = res->irq;
priv->wol_irq = res->wol_irq;
priv->lpi_irq = res->lpi_irq;
-
+
+ cpumask_clear(&cpumask);
+ cpumask_set_cpu(cpu_id,&cpumask);
+ irq_set_affinity(priv->dev->irq,&cpumask);
+
if (res->mac)
memcpy(priv->dev->dev_addr, res->mac, ETH_ALEN);
nice -n -20 iperf -c 192.168.30.100 -i 1 -t 5 -w 1M -d -P8
应该是这应用问题,我以前在屏蔽房用ixhhariot 跑是正常的, 电脑是XP系统, ixcharoit 是6.7版本, 板子上endpoint.apk version 7.30 SP1 build 32.
不记得是不是这个endpoint
使用:
3399以太网用ixCharoit 跑双流.zip
endpoint.apk
可以直接测好丢过去,硬要他们自己解自己系统资源问题,就怪他们
切换CPU core 中断验证。
例子
cat /proc/interrupts 查看相应中断号
echo 2 > /proc/irq/51/smp_affinity_list //把中断放到cpu2上执行
cat /proc/interrupts //查看记数
或者将eth0 的中断平均到各个CPU
diff --git a/drivers/irqchip/irq-gic-v3.c b/drivers/irqchip/irq-gic-v3.c
index 79b60aa..1e0fa6c 100644
--- a/drivers/irqchip/irq-gic-v3.c
+++ b/drivers/irqchip/irq-gic-v3.c
@@ -421,6 +421,12 @@ static void __init gic_dist_init(void)
affinity = gic_mpidr_to_affinity(cpu_logical_map(smp_processor_id()));
for (i = 32; i < gic_data.irq_nr; i++)
gic_write_irouter(affinity, base + GICD_IROUTER + i * 8);
+
+ /* irq 44 routed to all cpus */
+ affinity = affinity | 0x80000000;
+ gic_write_irouter(affinity, base + GICD_IROUTER + 44 * 8);
+ /* Decrease irq 44 priority value from default 0x0a to 0x09 */
+ writel_relaxed(0xa0a0a090, base + GICD_IPRIORITYR + rounddown(44, 4));
}
static int gic_populate_rdist(void)
@@ -658,7 +664,8 @@ static int gic_set_affinity(struct irq_data *d, const struct cpumask *mask_val,
reg = gic_dist_base(d) + GICD_IROUTER + (gic_irq(d) * 8);
val = gic_mpidr_to_affinity(cpu_logical_map(cpu));
- gic_write_irouter(val, reg);
+ if (gic_irq(d) != 44)
+ gic_write_irouter(val, reg);
/*
* If the interrupt was enabled, enabled it again. Otherwise,
31、delay的时间概念
具体的意义是相位差,推迟步进单位是2ns,最新回环补丁文档有介绍
可以理解为写delay让txc(clk)和txd(data)相位不断的靠近,具体效果还是需要用工具测量吞吐
32、添加以太网选项
redmine 249892提到本条redmine需要mid的以太网设置补丁,请参考附件10.0_ethernet_mid_settings.zip。
其中,device\rockchip\rk3399\下的宏开启补丁,如果是其他产品,请根据对应产品在device\rockchip\XXX\目录打上
补丁完全打上后,可以在settings > network&intent 看到 ethernet选项
33、PX30以太网配置
&gmac {
phy-supply = <&vcc_phy>;
assigned-clocks = <&cru SCLK_GMAC>;
assigned-clock-parents = <&gmac_clkin>;
clock_in_out = "input";
pinctrl-names = "default";
pinctrl-0 = <&rmii_pins &mac_refclk>;
snps,reset-gpio = <&gpio2 13 GPIO_ACTIVE_LOW>;
snps,reset-active-low;
snps,reset-delays-us = <0 50000 50000>;
status = "okay";
};
34、phy休眠
二级休眠是ping不通
如果phy也进行休眠,没有封包过来。主控是不会被唤醒的,也ping不通。目前这种情况属于正常。 我们服务器上是有个magic packet 唤醒主控的补丁,要用工具发魔术包去唤醒
35、以太网反复up down
mac clk不精
36、NAT转发
如果你是wlan0 上外网,eth1通过NAT共享wlan0,只需要加enableNat 就可以了,当然首先要保证pc 与3399 eth1可以ping 通。 不需要用脚本,直接看日志即可。
private void startDhcpServer() {
if (DBG) Log.d(TAG, "startDhcpServer");
String startIp = SystemProperties.get("persist.dhcpserver.start", "192.168.1.150");
String endIp = SystemProperties.get("persist.dhcpserver.end", "192.168.1.250");
String[] dhcpRange = {startIp, endIp};
try {
mNMService.tetherInterface(mIface);
mNMService.startTethering(dhcpRange);
+ mNMService.enableNat("eth1","wlan0");
37、静态IP设置后,没插网线控制phy启动状态carrier为0
hcq@ubuntu:~/33997.1/kernel/drivers/net/ethernet/stmicro/stmmac$ git diff stmmac_main.c
diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
index 58a490e..b14791b 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
@@ -1855,7 +1855,7 @@ static int stmmac_open(struct net_device *dev)
napi_enable(&priv->napi);
netif_start_queue(dev);
-
+ netif_carrier_off(dev);
#ifdef CONFIG_DWMAC_RK_AUTO_DELAYLINE
if (!priv->delayline_scanned) {
38、修改phy id
启动的第一步会去读phy id,可能有概率性获取不到的,先设置看是否有其它问题
--- a/drivers/net/phy/phy_device.c
+++ b/drivers/net/phy/phy_device.c
@@ -325,6 +325,9 @@ static int get_phy_id(struct mii_bus *bus, int addr, u32 *phy_id,
{
int phy_reg;
+ *phy_id = 0x11223344;
+ return 0;
+
if (is_c45)
return get_phy_c45_ids(bus, addr, phy_id, c45_ids);
&mdio1 {
rgmii_phy1: phy@0 {
- compatible = "ethernet-phy-ieee802.3-c22";
+// compatible = "ethernet-phy-ieee802.3-c22";
+ compatible = "ethernet-phy-id1234.d400", "ethernet-phy-ieee802.3-c22";
reg = <0x0>;
};
};
或者在属性上面设置
39、
单纯跑iperf , 查看一下CPU 频率
cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_cur_freq
多线程-P 4 确认吞吐有到940, 并查询CPU 频率。
cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_cur_freq
下面几个调频策略都试一下。
echo interactive > /sys/devices/platform/dmc/devfreq/dmc/governor
echo ondemand> /sys/devices/platform/dmc/devfreq/dmc/governor
echo performance> /sys/devices/platform/dmc/devfreq/dmc/governor
DDR测试
cat /sys/devices/platform/dmc/devfreq/dmc/available_frequencies
echo userspace > /sys/devices/platform/dmc/devfreq/dmc/governor
echo 800000000 > /sys/devices/platform/dmc/devfreq/dmc/min_freq
cat /sys/devices/platform/dmc/devfreq/dmc/cur_freq
cat /sys/devices/platform/dmc/devfreq/dmc/load
CPU
cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_available_frequencies
cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_available_governors
echo userspace > /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor
echo 1512000 > /sys/devices/system/cpu/cpu0/cpufreq/scaling_min_freq
cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_cur_freq
echo performance > /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor
调整 协议相关参数, tcp相关参数改大,再把iperf的window size设为2MByte后TX单线程就能跑到平均380M~392Mbps,较之前差不都有50-70M收益
echo 1048576 > /proc/sys/net/core/wmem_max
echo 1048576 > /proc/sys/net/core/rmem_max
echo "4096 1048576 1048576" > /proc/sys/net/ipv4/tcp_rmem
echo "4096 1048576 1048576" > /proc/sys/net/ipv4/tcp_wmem
echo 4193104 > /proc/sys/net/ipv4/tcp_limit_output_bytes
echo 1048576 > /proc/sys/net/ipv4/udp_rmem_min
echo 1048576 > /proc/sys/net/ipv4/udp_wmem_min
40
android7.1开始的kernel取消了之前版本的/proc/last_kmsg的节点,因此如果想要查看上一次的kernel log,可以在dts里面添加以下节点信息查看
diff --git a/arch/arm64/boot/dts/rockchip/rk3368-p9.dts b/arch/arm64/boot/dts/rockchip/rk3368-p9.dts
index bc0e0e2..7650d5f 100644
--- a/arch/arm64/boot/dts/rockchip/rk3368-p9.dts
++ b/arch/arm64/boot/dts/rockchip/rk3368-p9.dts
@ -70,6 +70,20 @
};
};
ramoops_mem: ramoops_mem {
+ reg = <0x0 0x110000 0x0 0xf0000>;
+ reg-names = "ramoops_mem";
+ };
ramoops {
+ compatible = "ramoops";
+ record-size = <0x0 0x20000>;
+ console-size = <0x0 0x80000>;
+ ftrace-size = <0x0 0x00000>;
+ pmsg-size = <0x0 0x50000>;
+ memory-region = <&ramoops_mem>;
+ };
+
sdio_pwrseq: sdio-pwrseq {
compatible = "mmc-pwrseq-simple";
clocks = <&rk818 1>;
130|root@rk3399:/sys/fs/pstore # ls
cd /sys/fs/pstore
dmesg-ramoops-0 上次内核panic后保存的log。
pmsg-ramoops-0 上次用户空间的log,android的log
ftrace-ramoops-0 打印某个时间段内的function trace
console-ramoops-0 last_log 上次启动的kernel log,但只保存了优先级比默认log level 高的log。
看是否有最后挂掉的日志。
41、io指令没有怎么办
加CONFIG_DEVMEM=y
CONFIG_DEVTMPFS_MOUNT=y
然后编译
42、rv1126 io 配成不匹配1.8V后恢复3.3要做的修改,参考重启慧居的doc
当改为3.3V后看到配驱动强度为LEVEL0/LEVEL1时,MAC_CLK的占空比已失真,需要配到LEVEL2才正常,软件默认为LEVEL1驱动强度。
43、IP175拔掉网线之后,以太网的图标还在
将CONFIG_IP175D_PHY 关闭
44、udp或tcp吞吐丢包
hcq@ubuntu101:~/RK3399/RK3399_LINUX_SDK_Release/kernel/drivers/net/ethernet/stmicro/stmmac$ git diff .
diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
index b3b8d87..df62d72 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
@@ -73,7 +73,7 @@ MODULE_PARM_DESC(phyaddr, "Physical device address");
#define STMMAC_TX_THRESH (DMA_TX_SIZE / 4)
-static int flow_ctrl = FLOW_OFF;
+static int flow_ctrl = FLOW_AUTO;
module_param(flow_ctrl, int, S_IRUGO | S_IWUSR);
MODULE_PARM_DESC(flow_ctrl, "Flow control ability [on/off]");
@@ -723,7 +723,7 @@ static void stmmac_adjust_link(struct net_device *dev)
priv->oldduplex = phydev->duplex;
}
/* Flow Control operation */
- if (phydev->pause)
+ //if (phydev->pause)
priv->hw->mac->flow_ctrl(priv->hw, phydev->duplex,
fc, pause_time);
@@ -827,6 +827,7 @@ static int stmmac_init_phy(struct net_device *dev)
priv->oldlink = 0;
priv->speed = 0;
priv->oldduplex = -1;
+ priv->flow_ctrl = FLOW_AUTO
if (priv->plat->phy_node) {
--- a/drivers/net/ethernet/rockchip/gmac/stmmac_main.c
+++ b/drivers/net/ethernet/rockchip/gmac/stmmac_main.c
@@ -3151,7 +3151,7 @@ struct stmmac_priv *stmmac_dvr_probe(struct device *device,
ndev->netdev_ops = &stmmac_netdev_ops;
- ndev->hw_features = NETIF_F_SG | NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM |
+ ndev->hw_features = NETIF_F_SG | NETIF_F_IPV6_CSUM |
NETIF_F_RXCSUM;
ndev->features |= ndev->hw_features | NETIF_F_HIGHDMA;
ndev->watchdog_timeo = msecs_to_jiffies(watchdog);
45、休眠给电
diff --git a/arch/arm64/boot/dts/rockchip/rk3399-evb-rev3.dtsi b/arch/arm64/boot/dts/rockchip/rk3399-evb-rev3.dtsi
index be7c60b..d15b088 100644
--- a/arch/arm64/boot/dts/rockchip/rk3399-evb-rev3.dtsi
+++ b/arch/arm64/boot/dts/rockchip/rk3399-evb-rev3.dtsi
@@ -315,7 +315,7 @@
regulator-boot-on;
regulator-name = "vcc3v3_s0";
regulator-state-mem {
- regulator-off-in-suspend;
+ regulator-on-in-suspend;
};
};
};
46、更新3568sdk出现获取不到ip问题
selinux的问题
adb shell setenforce 0
47、3568 4.19的内核没有对phy地址自动扫描功能造成读不到phy
rgmii_phy0: phy@0 {
compatible = "ethernet-phy-ieee802.3-c22";
- reg = <0x0>;
+ reg = <0x1>;
};
};
rgmii_phy0: phy@0 {
compatible = "ethernet-phy-ieee802.3-c22";
- reg = <0x1>;
+ reg = <0x0>;
};
};
phy地址设置0x0或者0x1都测试一下
48、关闭pause帧
ethtool -A eth0 tx off
ethtool -A eth0 rx off
ethtool -a eth0
Pause parameters for eth0:
Autonegotiate: on
RX: on
TX: on
ethtool -s eth0 speed 100 duplex full autoneg off
--------------rtl8211------------------
phy mac
xtal1 mac_clk&clkout(一起连,外部不接自激电路就直接mac_clk给,dts上设置output直接给xtal1,百兆网是25M,千兆网是125M)
tx0 - tx0
tx1 - tx1
tx2 - tx2
tx3 - tx3
txen- txen
txclk-txclk
rx0 - rx0
rx1 - rx1
rx2 - rx2
rx3 - rx3
rxen-rxen
rxclk-rxclk
rxdv-rxdv
mdio-mdio
mdc-mdc
GPIO - phy rst
注意每根线都得连上,少连一根都会找不到phy设备,还有rst也要看看有没有使能,如果吞吐有问题再调tx rx delay
---------------lan8720A---------------------
LED1_AD1 接下拉的时候 INTB才会输出CLK(接25M晶振,内部倍频 INTB输出50M)
DNP(do not populate)不焊接的意思,LED0_AD0接下拉是设置芯片地址为0
phy mac
xtal1 mac_clk
tx0 - tx0
tx1 - tx1
//tx2 - tx2
//tx3 - tx3
txen- txen
//txclk-txclk
rx0 - rx0
rx1 - rx1
//rx2 - rx2
//rx3 - rx3
//rxen-rxen
//rxclk-rxclk
rxdv-rxdv
rxer-rxer
mdio-mdio
mdc-mdc
gpio - phy rst