完善资料让更多小伙伴认识你，还能领取20积分哦， 立即完善>
3天内不再提示
电子发烧友论坛|
在和谐2.04与MZ DA系列芯片，我发现一个TCP套接字将无法恢复从半秒拔掉。这是在交换机上拔出的，所以设备不会丢失链接，只是从LAN到连接设备的连接。192.1680.100是服务器，192.1680.97是PIC。我不知道要找什么，但是它看起来有点像一个bug。重发从来没有停止过，也从未成功过。TCPIPpTCPUPUSTEADE返回0直到我杀死SoCKE.ACON.NET/Pux/PIC32.pCAP
以上来自于百度翻译 以下为原文 On hARMony 2.04 with the MZ DA series chip, I am finding that a TCP socket will not recover from a half second unplug. This is unplugged at the switch, so the device doesn't lose the link, just connection from the lan to connected device. 192.168.0.100 is the server, and 192.168.0.87 is the PIC. I am not sure what to look for, but it seems a bit like a bug perhaps. The retransmissions never stop and never succeed. TCPIP_TCP_PutIsReady returns 0 until I kill the socket. aercon.net/Public/PIC32.pcap 
0
|
|
相关推荐
14个回答
|
|
|
什么应用程序运行在192.1680.100作为服务器？在断开连接之后，服务器仍在运行和运行吗？一个WiReSARK日志会有所帮助。
以上来自于百度翻译 以下为原文 What application is running on the 192.168.0.100 as server? Is that server still up and running after the disconnect? A wireshark log will help. |
|
|
|
|
|
它是ICECAST2。我在第一篇文章中附加了一个链接到WiReSARK日志。我知道最初的问题是TCP/IP是围绕着It16构建的。到处都是溢出来的地方。如果你给一个TCP套接字提供足够的内存来做一些实际的工作，并且把这个大小设置为65535，你会马上发现有ACK和SEQS的问题。
以上来自于百度翻译 以下为原文 It is icecast2. I attached a link to the wireshark log in the first post. I know the initial problem, is that the TCP/IP was built around int16. There are overflows lots of places. If you give a tcp socket enough memory to do some real work, and set the size to 65535, you'll see right off that there are issues with acks and seqs. |
|
|
|
|
|
这个论坛不会让我发布*.h文件，并允许我拒绝任何其他内容。这也是开始执行窗口缩放的一个开始。我有点不清楚，如果我在PSKT & GT；HoleSize和所有正在发生的签名到无符号转换的情况下，都围绕着527到5315行。
以上来自于百度翻译 以下为原文 This forum won't let me post *.h files, and gives me permission denied for anything else. http://aercon.net/Public/tcp.zip This is what I have so far. It is a start at implementing window scaling as well. I am a bit unclear if I have done it right around lines 5277-5315 in regards to pSkt->HoleSize and all the signed to unsigned conversions that are happening. |
|
|
|
|
|
使5176-5315甚至不能编辑，更不用说实际上传文件了。
以上来自于百度翻译 以下为原文 Make that 5176 - 5315 Can't even make edits, let alone actually upload files. |
|
|
|
|
|
你使用破损的链式图标来打破你试图编辑的帖子中的链接吗？
以上来自于百度翻译 以下为原文 Did you use the broken chain icon to break the link in the post you tried to edit ? |
|
|
|
|
|
我不知道，但我不这么认为。这是个巧妙的问题吗？-）
以上来自于百度翻译 以下为原文 I don't know, but I don't think so. Is that a trick question? :-) |
|
|
|
|
|
|
|
|
|
|
|
这是你正在使用的TCP的修改版本还是标准的和声分布？
以上来自于百度翻译 以下为原文 Is this a modified version of TCP that you're using or the standard Harmony distribution? |
|
|
|
|
|
我从2.04开始。我所连接的是我在那个堆栈上修改的地方，接近我需要去的地方。在没有修改的情况下，它在我的情况下工作得很差。我认为如果设置TCP套接字大小大于0x7FFF，因为堆栈溢出，堆栈会被弄乱。如果您查看WiReSARK捕获，您可以看到SEQ的值减少了65535。这个捕获是用标准的2.04版本的TCP、C和TCP。
以上来自于百度翻译 以下为原文 I started with the 2.04. What I linked to is what I have modified on that stack to get close to where I need to go. With no modifications, it works rather poorly in my situation. I think the stack messes up if you set the tcp socket size larger than 0x7FFF because of overflows. If you look at the wireshark capture, you can see that the seq's get off by 65535. That capture was with the standard 2.04 version of tcp.c and tcp.h |
|
|
|
|
|
我把我的beta测试的MODS连接到了TCP、C和TCP。这似乎也加快了HTTP服务器的速度。至少如果你给他们足够的RAM。
以上来自于百度翻译 以下为原文 I attached my beta tested mods to tcp.c and tcp.h This seems to really speed up the http server as well. At least if you give them enough ram. Attachment(s) TCP_Large.zip (54.52 KB) - downloaded 22 times |
|
|
|
|
|
我已经用TX和RX缓冲器测试了16 KB，我发现了一些问题，正如你所报道的。这个问题已经确定了，它将是下一个版本的一部分。如果你需要的话，我可以给你发一个更新的副本。不管怎样，从IPEF和其他传输测试结果来看，一旦你过去了。16 kb的缓冲区，吞吐量的差别不够大，例如，使用（64 kb - 1）缓冲区而不是16 kb。
以上来自于百度翻译 以下为原文 I've performed some tests with TX and RX buffers > 16 KB and I've found some problems, as you reported. The issue has been fixed and it will be part of the next release. I can send you a copy with the updates, if you need it. Anyway, from the iperf and other transfer tests results, once you go past 16 KB buffers, the difference in throughput is not significant enough to justify, for example, working with (64 KB - 1) buffers instead of 16 KB. Thank you for finding and reporting this. |
|
|
|
|
|
我想我得把我的稿件继续写下去，不过还是要谢谢你。我添加了窗口缩放和32位大小的缓冲能力，我希望和声可以添加。DA系列可以为TCPIP模块使用未经缓存的DDR空间，它允许每件测试都能很好地工作。我的应用是音频流，所以它需要一些缓冲空间才能正常工作。
以上来自于百度翻译 以下为原文 I suppose I'll have to keep my copy going forward, but thanks anyway. I added window scaling and 32 bit size buffer capability which I wish harmony could add. The da series can use uncached ddr space for the tcpip module and it allows things to work pretty nice per my testing. My application is audio streaming so it needs some buffer space to work well. |
|
|
|
|
|
至于IPEF测试，您确实应该尝试在10MB连接到远程服务器上，看看您发现了什么。
以上来自于百度翻译 以下为原文 As for iperf testing, you really should try it on a 10mb connection to a remote server and see what you find. |
|
|
|
|
|
同意，窗口缩放是必要的，特别是在有外部资源的部分，它是开发计划的一部分。
以上来自于百度翻译 以下为原文 Agreed, window scaling is needed, especially on parts with external resources and it is part of the development schedule. |
|
|
|
|
只有小组成员才能发言，加入小组>>
MPLAB X IDE V6.25版本怎么对bootloader和应用程序进行烧录
473 浏览 0 评论
5793 浏览 9 评论
2334 浏览 8 评论
2224 浏览 10 评论
请问是否能把一个ADC值转换成两个字节用来设置PWM占空比？
3530 浏览 3 评论
1122浏览 1评论
有偿咨询，关于MPLAB X IPE烧录PIC32MX所遇到的问题
1095浏览 1评论
我是Microchip 的代理商，有PIC16F1829T-I/SS 技术问题可以咨询我，微信：A-chip-Ti
872浏览 1评论
MPLAB X IDE V6.25版本怎么对bootloader和应用程序进行烧录
473浏览 0评论
/9 
关注我们的微信
下载发烧友APP
电子发烧友观察
版权所有 © 湖南华秋数字科技有限公司
电子发烧友 （电路图） 湘公网安备 43011202000918 号 电信与信息服务业务经营许可证：合字B2-20210191
工商网监
湘ICP备2023018690号
1070
评论
淘帖
2878
