**** BEGIN LOGGING AT Sat Oct 06 03:00:01 2018 Oct 06 08:08:59 Hi, I'm getting this on latest snapshot with kernel version 4.14: Oct 06 08:09:03 possible DNS-rebind attack detected: internal-ambassador-production-v2-vpc-elb-2 Oct 06 08:09:04 051440118.us-east-1.elb.amazonaws.com Oct 06 08:11:10 with very slow amazon website Oct 06 08:11:40 lots of timeouts Oct 06 09:35:37 hi, I'm detecting on the armada-370 (mvebu target) net driver some mistakes, or I don't really understand this stuff Oct 06 09:35:50 i.e. https://github.com/torvalds/linux/blob/master/drivers/net/ethernet/marvell/mvneta.c#L1374 Oct 06 09:36:21 it says bit 16 but then it should be 0x10000 not 0x20 Oct 06 09:36:47 and accordingly to armada370 datasheet is indeed the bit16 Oct 06 09:38:29 also some register offsets definitions doesn't seem ok Oct 06 09:38:33 https://github.com/torvalds/linux/blob/master/drivers/net/ethernet/marvell/mvneta.c#L241 Oct 06 09:39:00 I don't find any mtu with that offset in the datasheet Oct 06 12:11:00 nbd: I am now getting this warnings from kconfig: https://pastebin.com/Mf1DdeN0 Oct 06 12:26:42 pushed a fix Oct 06 13:03:12 Hi, how can I debug s/w flow offload issues? losing a lot of packets with it turned on. Logs don't show anything Oct 06 13:18:25 shalzz: what kind of packets are you losing, how are you reproducing the issue? Oct 06 13:20:23 nbd: just browsing the net and visiting websites. Some of them load others just continuously time-out Oct 06 13:20:49 what device are you testing on? Oct 06 13:20:51 Turning off offload fixes these issues Oct 06 13:21:16 Tl-wr841n v11 Oct 06 13:24:21 do you get any stale entries with the OFFLOAD flag set and high timeout in /proc/net/nf_conntrack? Oct 06 13:24:25 can you show me your network config? Oct 06 13:26:38 nbd: https://paste.gnome.org/pjtoywzrk Oct 06 13:27:13 trying with OFFLOAD set again Oct 06 13:30:40 I'm getting some TIME_WAIT in /proc/net/nf_conntrack Oct 06 13:32:13 is it always the same websites that time out with offload enabled? Oct 06 13:33:20 yup Oct 06 13:33:23 this is what I have so far: https://paste.gnome.org/pbgeizeh8 Oct 06 13:35:29 please do a tcpdump on pppoe-wan (saved to a .pcap file) with just one access to a website that's broken with offload Oct 06 13:35:40 once with offload enabled, and once with offload disabled Oct 06 13:35:49 then please send me those files Oct 06 13:35:54 so i can compare and see if i can figure out what's wrong Oct 06 13:41:45 Ok compiling tcpdump in and send you those files Oct 06 13:47:53 oops can't fit tcpdump Oct 06 13:48:42 I'll need to remove some packages Oct 06 13:49:17 shalzz: Did you try with tcpdump-mini? Oct 06 13:50:39 yup, Oct 06 13:50:48 too big by 95180 bytes Oct 06 13:53:10 removed luci-app-upnp hopefully that saves 95K Oct 06 14:06:04 shalzz: i'm heading out now, you can send them to nbd@nbd.name Oct 06 14:06:11 will take a look at them when i have time Oct 06 14:06:12 thanks Oct 06 17:53:22 i've got a tp-link archer c7 v5. running openwrt snapshot. if i use the wifi extensively (speedtest) the router crashes and reboots after a few seconds no matter if i use 2,4ghz or 5ghz wifi. the crash leaves the following in dmesg: https://gist.github.com/rotanid/401ec680f74e01048cf759d34f94ce0c Oct 06 17:54:07 is it true the v5 does not have detachable antennae? Oct 06 17:54:24 yes Oct 06 17:54:36 saving another few cents in production... Oct 06 19:00:23 neoraider: I ran into a problem with your commit in the wireless-regdb: https://git.kernel.org/pub/scm/linux/kernel/git/sforshee/wireless-regdb.git/commit/?id=f3c4969c2485e7d0935041c50ce5f69894a0e7ec Oct 06 19:00:42 my python installation does not have the bytes module Oct 06 19:05:55 hmm do we want to make users install python-future ? Oct 06 21:02:18 Hi - if anyone is around to review PR #7122, that would be great -I'll be online for a while. Oct 06 21:20:04 Hauke: huh, I wasn't aware that that wasn't always included... I think it would make more sense to require both Python2 and Python3 for the moment (with the idea to kill Python2 as soon as possible - the biggest roadblocks being uboot (at least for some targets) and the Layerscape LSDK) Oct 06 21:23:22 And IIRC the LSDK only has a single Python script that should be easy to port to Python3 Oct 06 21:24:22 The u-boot Python dependency is a mess... some targets want python-dev, so it can bulid native extensions (but I think we patch that away for now where possible?) Oct 06 21:25:51 neoraider: some uboot targtes wanted swig, but I patched it away Oct 06 21:25:59 it is needed for binman Oct 06 21:26:08 or something called similary Oct 06 21:28:10 I looked at the issue a few months ago... according to the copyright comments, the binman scripts are only like 2~3 years old, it's really annoying that they added legacy python scripts when most projects are trying to upgrade to Python3 ASAP Oct 06 21:29:10 I didn't noticed they use python2 only Oct 06 21:29:19 I just wanted to get rid of it ;-) Oct 06 21:29:34 Hauke: I've tested your WPA3 patches (still with the old 4.14 based backports because of WDS and ath10k), the first impression is positive with using "option encryption 'psk3-mixed'" (which I think is supposed to do the right thing for wpa2-psk/ccmp and wpa3) as AP - I haven't quite found out how to configure a client connection with wpa3 yet, but I only guessed "psk3" and "psk-sae" so far Oct 06 21:30:08 OpenEmbedded avoids the whole issue by building native Python 2 and 3 itself, rather than making it a prerequisite Oct 06 21:33:53 pkgadd: I used wpa_supplicant master on the client to conenct Oct 06 21:34:47 by default openwrt now forces ieee80211w when configiured to sae as this is in the WPA3+SAE starand now Oct 06 21:35:11 sadly the wifi FW of my intel 6300 card card in some of my device s Oct 06 21:35:11 yep, that will be my next test before looking deeper into the OpenWrt/ WDS-AP <-- wpa3 WDS link --> OpenWrt WDS-CLient/ repeating the signal Oct 06 21:35:24 sadly the wifi FW of my intel 6300 card does not support ieee80211w Oct 06 21:36:14 neoraider: we could also build python, how long does it take? Oct 06 21:36:35 Hauke: not sure, let me check... Oct 06 21:37:56 I had 802.11w enabled before already (nbg6817/ QCA9984 and tl-wdr3600/ AR9344), so 802.11w shouldn't be an issue - but I'll do further tests on Debian and AR5007EG/ ath5k next (also with working 802.11w); ipw2200 (ancient, I know) won't do 802.11w either Oct 06 21:37:58 Hauke: that way we could also make python-dev available, hopefully avoiding the need to patch binman out of u-boot Oct 06 21:39:46 neoraider: yes just building something is much easier than havng a long list of prerequists Oct 06 21:42:53 Hauke: about 2:40 minutes on my dualcore CPU for Python3 (including other host deps that weren't already built) Oct 06 21:43:56 So build time shouldn't be a problem, the only issue I see is the complexity of the current Python package makefiles Oct 06 21:50:07 google doesn't like 802.11w either for some reason Oct 06 21:50:22 my nexus 6p and chromecast do not support it Oct 06 21:52:50 most of the reasons are the relatively old wpa_supplicant and kernel(s) on android, apparently android 9 is supposed to support wpa3 (and with that 802.11w) - if there is a recent enough kernel (at least 3.8, which sounds easy, but...) Oct 06 21:54:19 the nexus 6p won't get android 9, so I need to get a new phone Oct 06 21:54:27 it's almost at eol Oct 06 21:54:47 doesn't the nexus 6p have lineageos support? Oct 06 21:55:01 probably, but I've never looked into it Oct 06 21:55:24 but yes, the general update situation around android could be a *lot* better Oct 06 21:56:15 apple is actually not that bad with regards to supporting their devices, but I still like android since it's a bit more open Oct 06 21:56:38 I have one device that has a decent chance for android 9 via lineageos (currently already running lineageos 15.1), but kernel 3.4 might spoil wpa3 support Oct 06 21:57:32 the others almost definately won't Oct 06 21:58:53 I might give it a try if I decide that buying a new phone is too expensive (which is it these days) Oct 06 22:01:39 sae does not need any HW or FW support, you only need a recent wpa_suplicant Oct 06 22:02:11 but WPA3+sae requires 802.11w Oct 06 22:02:41 only for WPA2 it is optional, but I think for the current WPA2 certifiation you also have to support ieee80211w Oct 06 22:03:39 does that mean a certification from the wifi alliance? Oct 06 22:04:40 yes Oct 06 22:05:20 the issue is that major companies don't really seem to care about that certificaion Oct 06 22:05:22 I think I read somewhere that ieee80211w support is now required for WPA2 certification Oct 06 22:05:30 you do not have to force this, but supprot it Oct 06 22:05:36 for WPA3 clients you have to force it Oct 06 22:06:02 I think you need it to call your product wifi Oct 06 22:06:26 the last iphone to have a wifi alliance certification was the iphone 4s Oct 06 22:08:37 but it will be in their interest to impement wpa3 sooner or later, and as long as it works that's fine by me Oct 06 22:10:08 apple has to save some money ;-) Oct 06 22:11:00 if they want apple can probably support wpa3 in all their devices Oct 06 22:11:25 that probably won't happen with devices that are eol Oct 06 22:13:17 anyhow, I'll be on my way for now Oct 06 22:13:19 gn **** ENDING LOGGING AT Sun Oct 07 02:59:59 2018