loshia

Thank You! Best

Hi Sfstudio, Can i ask you if spinlock instead write_lock_bh can be used in linux-2.6.21.x/drivers/net/pppoe.c to speed (if it will speed it up at all??) up pppoe also like the following commit: 18:41 sfstudio committed 9be9534 VPN pptp/l2tp drivers now use spinlock instead write_lock_bh(speedup). 10X

Thanks yusuff!

Sfstudio, ----- If vlan_double_tag=0 use hardware vlan offload. --------- сделать nvram_set vlan_double_tag 1 && reboot, но тогда отключиться аппаратный оффлоад вланов -------- My Russian is not so good but you are saying two opposite things. which of two is right? Probably English one? 10X!

Sfstudio, There is a problem with 802.1q vlan support (latest git from today) because of MAX_PACKET_SIZE - 1518 (1522 is required for vlan) defined in raether.h raether.c if (rx_skb->len > MAX_PACKET_SIZE) { if (net_ratelimit()) ei_local->stat.rx_dropped++; printk("ERROR: frame is BIG !!!\n"); bReschedule = 1; break; } Moreover When ERROR: frame is BIG !!! appears in console it seems for me that router freezes or at least there is something wrong with error handling in that case. For me solution was to comment that part of code. i am not pretedning that commenting out is the right approach but it works for me. Just sharing that bug... Best

Hello Sfstudio, I know that i am bothering you too much but with Acorp_WR-150N-1T1R.1.10.12.RU.30092011.bin i found that AuthMode=WPAPSKWPA2PSK is not working anymore Quick info - Laptop has Intel pro wireless 5100 with latest drivers updated What is happening that laptop connects just fine. 1. After couple of power off/on events of the router i keep getting flowing every seccond ~ # ASSOC - Assign AID=1 to STA 00:21:5d:a7:c3:6e ASSOC - 11n HT STA ASSOC - Assign AID=1 to STA 00:21:5d:a7:c3:6e ASSOC - 11n HT STA ASSOC - Assign AID=1 to STA 00:21:5d:a7:c3:6e ASSOC - 11n HT STA And laptop WiFi status says Identifying - and does not gets any IP I have WiFi debug version with detailed logs (i can send them if you want) but in short there is a problem with WPA and Intel wifi adapters I keep getting (ap_wpa.c) DBGPRINT(RT_DEBUG_TRACE, ("WPARetryExec::MSG1 timeout, pEntry->ReTryCounter = %d\n", pEntry->ReTryCounter)); I can assure you that it was not happening with oldest versions. More over if laptop connects without powering of the router after some time it looses wifi again. Do you have any idea what can cause this? 10X Best Short Logs ####### Send L2 Frame Mac=00:21:5d:a7:c3:6e ASSOC - Assign AID=1 to STA 00:21:5d:a7:c3:6e ASSOC - legacy STA ASSOC - legacy STA ASSOC - MODE=0, MCS=0 AuthMode=7, WepStatus=4, WpaState=7, GroupKeyWepStatus=4 WMMCapable=1, Legacy AGGRE=0, PiggyBack=0, RDG=0, TxAMSDU=0, IdleTimeout=300 Enqueue EAPoL-Start-PSK for sta(00:21:5d:a7:c3:6e) WpaEAPOLStartAction ===> PortSecured(2), WpaState(7), AuthMode(7), PMKID_CacheIdx(255) ===> WPAStart4WayHS ===> ConstructEapolMsg for WPA2 Pairwise Message 1 Body length = 95 Key length = 32 APHardTransmit --> clear eap frame !!! <=== WPAStart4WayHS: send Msg1 of 4-way Receive EAPOL-Key frame, TYPE = 3, Length = 119 WpaEAPOLKeyAction ===> Receive EAPoL-Key frame from STA 00-21-5D-A7-C3-6E ===> PeerPairMsg2Action RTMPGetTxTscFromAsic : WCID(212) TxTsc 0x00-0x01-0x00-0x00-0x00-0x00 RTMPInsertRSNIE : The length is PMKID-List is invalid (0), so don't insert it. ===> ConstructEapolMsg for WPA2 Pairwise Message 3 Body length = 157 Key length = 32 APHardTransmit --> clear eap frame !!! <=== PeerPairMsg2Action: send Msg3 of 4-way WPARetryExec---> ReTryCounter=11, WpaState=9 WPARetryExec::MSG1 timeout, pEntry->ReTryCounter = 11 Send DEAUTH frame with ReasonCode(15) to 00:21:5d:a7:c3:6e

Thanks, I got the idea. I will disable (not compile) hw_nat at all and see if 65000 bytes are ok. In general what you are explaining is that hw_nat do not have enough buffer to handle big packets. Any way thank you for the info:) Best Update: Actually hw_nat hooks are inside eth driver, netfilter and a couple of other places. Without hw_nat compiled (ping 65000 bytes are ok. Performance PPPOE + NAT router makes 10.3-11 MB download which makes about 80-88 Mbps download. More over browsing goes smoothly in my personal opinion. Drawbacks - high soft irq load, probably bidi traffic decreases also (not tested) Thank you very much for heading me what can cause such behavior. As long as i do not need to safe irq resources for the moment and 80 MBPS is just perfect for me, my personal choice is to disable hw_nat, because what i care is not to max out 100 Mbs UTP (which is not possible in real world), but to have smooth browsing without delay and jitters

Sfstudio, Thank you for your reply. I know quite well that 1500 bytes is max packet size (Without fragmentation). I am tiring to say that packet fragmentation is broken (in the router) which is affecting the performance of the router. Back to your reply... Same thing happens when hw_nat and soft nat are not used. (hw_nat mod unloaded) + soft nat (/proc/sys/...fasnat=0) I am almost sure that PMTU discovery is on. But anyway i am just wondering why 65000 bytes used to work with old revisions when hw_nat and fast nat were off? Do you mean that just compiling hw_nat as module (even though not loaded) limits icmp packet size? Again just to make things easier what config must be done with latest build so that router can pass 65000 bytes when both nat + FW disabled. I know that you are too busy but if you spend a couple of mins to test it you will see that there is no way to ping WAN or LAN from a wired station with 65000 bytes. Thank you very much in advance PS: please understand me correct i am not trying to make you mad or nervous but as of my knowledge something is wrong and that is why i am posting again:)

Sfstudio, arp is working fine with Acorp_WR-150N-1T1R.1.10.12.RU.30092011.bin Thank you!! But unfortunately with same revision (probably a couple of revs before) icmp fragmentation seems to be broken again. If you remember a long time ago it was broken again and you fixed it. As you said in Russian the router suffered it's "performance" because of that. Probably it suffers again The result ping the router from wired PC (connected to LAN) with big packets C:\Documents and Settings\H>ping 192.168.1.1 -l 30000 Pinging 192.168.1.1 with 30000 bytes of data: Reply from 192.168.1.1: bytes=30000 time=6ms TTL=64 Reply from 192.168.1.1: bytes=30000 time=6ms TTL=64 Reply from 192.168.1.1: bytes=30000 time=6ms TTL=64 Reply from 192.168.1.1: bytes=30000 time=6ms TTL=64 Ping statistics for 192.168.1.1: Packets: Sent = 4, Received = 4, Lost = 0 (0% loss), Approximate round trip times in milli-seconds: Minimum = 6ms, Maximum = 6ms, Average = 6ms C:\Documents and Settings\H>ping 192.168.1.1 -l 50000 Pinging 192.168.1.1 with 50000 bytes of data: Request timed out. Request timed out. Request timed out. Request timed out. Ping statistics for 192.168.1.1: Packets: Sent = 4, Received = 0, Lost = 4 (100% loss), The result is absolutely the same when pinging router WAN interface (PC is connected in same switch with WAN of the router) The interesting part is that Wireless pc connected to ra0 can ping LAN (br0 - 192.168.1.1 and WAN IP of the router with 65000 bytes without issues) but can not ping nothing in the WAN subnet with bigger packet than 29000-30000 bytes What can cause this? Thanks!

Sfstudio, I have upgraded my router to Acorp_WR-150N-1T1R.1.10.11.RU.24092011.bin. I had been using net.ipv4.conf.all.arp_ignore = 1 with oldest versions to avoid LAN interface responding to arp requests coming from WAN interface. According to kernel docs arp_ignore - INTEGER Define different modes for sending replies in response to received ARP requests that resolve local target IP addresses: 0 - (default): reply for any local target IP address, configured on any interface 1 - reply only if the target IP address is local address configured on the incoming interface What I did is: 1. I upgraded the router to Acorp_WR-150N-1T1R.1.10.11.RU.24092011.bin 2. fs cleanall 3. ping 192.168.1.1 – from my pc OK 4. sysctl –w net.ipv4.conf.all.arp_ignore=1 (inside router CLI) 5. arp –d 192.168.1.1 (from My PC) 6. ping 192.168.1.1 – from my pc and I got core dump UPDATE: i tracked back changes and changing ipv4/arp.c line 414 from return !inet_confirm_addr(in_dev, sip, tip, scope); to return !inet_confirm_addr(dev, sip, tip, scope); Solved the issue at least for my build. I do not know if this is right or wrong and what side effects might appear If you could fix that it would be great! Thank You in advance! Here is the kernel dump # Unhandled kernel unaligned access[#1]: Cpu 0 $ 0 : 00000000 80240000 00000003 00000001 $ 4 : 00000003 00000000 0101a8c0 000000fe $ 8 : 81fe9831 00000000 80245d84 80240000 $12 : 7fff6380 00000000 00000400 7d88612b $16 : 81fe6e60 81975000 81ce7880 8190a820 $20 : 81fe982c 00000040 801f0000 80200000 $24 : 00000000 8018bb80 $28 : 801ec000 801edc48 801ede20 8018b448 Hi : 00000000 Lo : 00000000 epc : 8018de24 Tainted: P ra : 8018b448 Status: 11008003 KERNEL EXL IE Cause : 00800010 BadVA : 0000000f PrId : 0001964c (MIPS 24K) Modules linked in: hw_nat(P) rt2860v2_ap nf_nat_ftp nf_conntrack_ftp Process swapper (pid: 0, threadinfo=801ec000, task=801ee188) Stack : 81fe6e60 81975000 81ce7880 81fe9824 81fe982c 8018b908 81d19780 ffffffff 801d6ef0 80200000 81975000 c01e5d30 81953400 81fe6c20 3701a8c0 0101a8c0 81ce7880 c006f000 00020000 c0190000 c008d1e0 c008d1f4 c008d260 c0200000 c0217990 c0189530 81953400 81fe6c20 819753a0 00000000 c00912b8 00000040 c008d298 81fe6e60 81955800 819753a0 00000000 81fe6e60 00000040 00010000 ... Call Trace:[<c01e5d30>][<c0190000>][<c0200000>][<c0189530>][<c0189720>][<c01d44c4>][<80008558>] Code: 3c028024 08063781 00000000 <8c83000c> 1060003e 00002021 00001021 14400037 00000000 Kernel panic - not syncing: Fatal exception in interrupt Rebooting in 1 seconds..

Kernel Linking problem Sftudio, I am not an expert and what i will post is the problem as of my knowledge with help of the Google of course On some pint onward – when the kernel size increases vmlinux can not be linked. For instance I am trying to compile IPv6 enabled kernel. All is fine until linking Here is what happens…. --------------------------- GEN .version `CHK include/linux/compile.h SKIPPED include/linux/compile.h LD vmlinux /root/wive/toolchain/bin/mipsel-linux-uclibc-ld: section .init.text loaded at [0000000080220000,000000008023545f] overlaps section .data loaded at [0000000080214000,000000008022313f] /root/wive/toolchain/bin/mipsel-linux-uclibc-ld: section .data.cacheline_aligned loaded at [0000000080224000,00000000802260bf] overlaps section .init.text loaded at [0000000080220000,000000008023545f] make[1]: *** [vmlinux] Error 1 make[1]: Leaving directory `/root/wive/linux-2.6.21.x' make: *** [linux] Error 1 ---------------------- From the other hand when I do decrease kernel size (disabling some of the kernel features) it links just fine. That is why I am thinking that it is related wit build scripts and kernel size. I can post you my kernel config if you want it of course Can you be so kind to comment what is wrong? PS: I am trying to build ipv6 kernel for 4M Flash....It may be connected I just figured out that compiling same kernel config with "hacked" # CONFIG_RT2880_FLASH_4M is not set CONFIG_RT2880_FLASH_8M=y and CONFIG_MTD_PHYSMAP_LEN=0x800000 Compiles just fine and produced image with size of 3.45Mb (i have disabled all user applications samba,transmission and so on) which is absolutely OK to fit in 4M flash But unfortunately device can not boot because ralink-flash.h and probably others are ***ed up (all fs partition sizes are wrong) 10X

~ # sysctl -a | grep doub net.ipv4.vlan_double_tag = 0 set 1. Super! works like a charm!!! Thank you very much Best

Sftudio, Vlan 802.1q support seems to be broken at least with fw rev => 1.8.8.RU.01092011 Rev 1.10.3.RU.13092011 was tried also. What happens is (1.10.3.RU.13092011): # vconfig add eth2.2 200 ~ # ifconfig eth2.2.200 up ~ # ip a a 10.11.12.13/24 dev eth2.2.200 ~ # iptables -L Chain INPUT (policy DROP) target prot opt source destination ACCEPT all -- anywhere anywhere ACCEPT all -- anywhere anywhere Chain FORWARD (policy DROP) target prot opt source destination ACCEPT all -- anywhere anywhere Chain OUTPUT (policy ACCEPT) target prot opt source destination ACCEPT all -- anywhere anywhere # ping 10.11.12.1 PING 10.11.12.1 (10.11.12.1): 56 data bytes --- 10.11.12.1 ping statistics --- 3 packets transmitted, 0 packets received, 100% packet loss eth2.2.200 Link encap:Ethernet HWaddr 00:0B:2B:77:1E:3A inet addr:10.11.12.13 Bcast:0.0.0.0 Mask:255.255.255.0 UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1 RX packets:0 errors:0 dropped:0 overruns:0 frame:0 TX packets:3 errors:0 dropped:0 overruns:0 carrier:0 collisions:0 txqueuelen:0 RX bytes:0 (0.0 B) TX bytes:138 (138.0 B) Actually interface sends 802.1q frames just fine (checked with tcpdump) but it can receive nothing. I was doing vlans from a long time ago so my setup is pretty fine. As long as you can see it is FW issue (or at least I think so), with oldest FW revisions VLAN was working just fine…Unfortunately I am not able to tell you which revision broke it. Can you fix that? Thank you in advance

first of all, author of dhcpv6 porting is lly, not me, second - there's no garantee that dhcpv6 will be included, but we will try, of course p.s а чо не на мове? :D Thank you for the answers! I understood two things: 1. sfstudio is quite busy because of some Chinese bugs flowing around 2. At some point bb might have dhcpv6 statefull client with PD - but it is not sure... So if someone give me a hint how to compile libstsdc++ (so i can play with dibbler) in current wive toolchain will be great! Thank you everybody in advance!!! best

Hey, Thank you for the quick reply. Are these patches going to support DHCP-PD (prefix delegation) - from the busybox ipv6 dhcp client? I tried to compile dibbler (g++) but got no luck. It requires libstdc++ which is not build by the toolchain. I need dibbler client only. Probably it does not require libstdc++. Can you give me quick hint how to recompile toolchain with libstdc++ enabled? Thank you very much in advance