**** BEGIN LOGGING AT Thu Jun 14 03:00:04 2018 Jun 14 04:55:45 nbd: current release for 18.06.0-rc1 is planned for 17th, do you think you can backport the flow offloading fixes by then? Jun 14 04:56:00 thats about three days from now Jun 14 04:56:43 we can tag on the evening of 17th and have the binaries uploaded during the night Jun 14 07:26:53 jow: yes Jun 14 07:29:09 nice updated mwlwifi driver Jun 14 07:29:38 hmmm I should try flow offloading again Jun 14 07:29:50 jow: seems that udp still needs fixing and my offload fixes work for tcp Jun 14 07:29:54 last time I tried it was working even with my IPsec+VTI setup Jun 14 07:29:55 will backport them now Jun 14 07:30:05 but my router started spontaneously rebooting Jun 14 07:30:17 without any crash dump on serial console :( Jun 14 07:47:48 * ldir starts another day of banging head against the ATM brickwall (along with mkresin) At least I think it's a different brick today. Jun 14 08:27:03 How does one get additional privs on flyspray? Working on FS#1567 and can't even mark it confirmed. Jun 14 08:29:23 you're working for mauro now :^) Jun 14 08:30:40 I've spent £130 for Mauro Jun 14 08:31:41 To get a phone line & ADSL service installed of the type he reported trouble on. Jun 14 08:32:54 You have no idea how much additional logging (printk) we've put in the kernel & pppd to try to work out WTF is going on. Jun 14 08:52:04 stintel: hm, why don't those get covered by the normal Content-Encoding:gzip and auto-decompression? Jun 14 08:52:40 ldir: you have it running now? Jun 14 08:53:43 dwmw2_gone: you must be f* joking Jun 14 08:53:44 did you try reverting c55fa3cccbc2c672e7f118be8f7484e53a8e9e77 as I suggested? Jun 14 08:53:57 I meant the line up, not the problem fixed Jun 14 08:54:18 seriously, we're so far down the rabbit hole! Well beyond that Jun 14 08:54:18 It works in 4.9 and not 4.14, right? Jun 14 08:54:23 correct Jun 14 08:54:37 and ATM *is* actually working; you can do br2684 Jun 14 08:56:56 dwmw2_gone: I have no idea, just that's how I got it working again Jun 14 08:57:13 if br2684 is working then you can send and receive ATM frames Jun 14 08:57:37 yes atm is working. Jun 14 08:57:52 so is this *really* lantiq-only? Jun 14 08:57:55 yes you can create an atm bridge and run pppoe over that Jun 14 08:58:40 what's really interesting is you can use 'pppoa' but with llc encaps AND IT WORKS. But only 'cos BT strip the llc stuff. Jun 14 08:59:32 You mean BR2684 Ether-over-ATM? Jun 14 08:59:34 if you use vc-mux then it stops. We see the request for CHAP session init, and pppd replies.... but somewhere, somehow, that reply goes missing. Jun 14 08:59:37 and then PPPoE (oA) ? Jun 14 09:00:14 you can do that. Jun 14 09:01:14 but without creating the bridge, you can set 'pppoa' as the protocol type, 'llc' as the encaps on vpi/vci 0.38 and that works too. Jun 14 09:01:26 ah, ok Jun 14 09:02:10 do we have reason to believe this is *only* lantiq? Jun 14 09:02:15 the difference from a pppd perspective is a 4 byte header Jun 14 09:02:28 except it isn't pppd, right? That's in the ATM layer. Jun 14 09:02:35 but... really ought to be in the ATM layer, not hardware-specific Jun 14 09:02:48 I am unaware of any directly support adsl on non-lantiq Jun 14 09:02:57 should I send you a Solos PCI card? Jun 14 09:03:27 or a USB ASDL modem would suffice, if it's generic linux ATM Jun 14 09:03:47 if you use vc-mux then it stops. We see the request for CHAP session init, and pppd replies.... but somewhere, somehow, that reply goes missing. Jun 14 09:04:00 So... you see *incoming* frames, but you aren't sure you're ever sending anything? Jun 14 09:04:44 and you've tried reverting c55fa3cccbc2c672e7f118be8f7484e53a8e9e77 ? Jun 14 09:05:10 we don't even see vcc_sendmsg being called Jun 14 09:06:00 and we tweaked the driver to be c55fa3cccbc2c672e7f118be8f7484e53a8e9e77 compliant.... before realising that it doesn't even get called :-) Jun 14 09:06:06 heh Jun 14 09:09:47 but that's a straight write() or send() from userspace on an ATM socket opened by pppd's pppoatm plugin, right? Jun 14 09:09:58 ditch pppd, just open the socket and try reading/writing? Jun 14 09:13:27 https://pastebin.com/Ze3bdmNw Jun 14 09:20:02 vs https://pastebin.com/zMEgaxEr - the difference in /etc/config/network is option encaps 'llc' (instead of vc) Jun 14 09:30:27 [ 1770.456434] pppoatm_send: c0 21 09 00 00 08 9f 6e a7 ed .!.....n.. Jun 14 09:30:27 [ 1770.464323] Jun 14 09:30:27 [ 1770.465739] __ppp_channel_push: put the packet back and try again later Jun 14 09:31:00 you basically never send that one? Jun 14 09:31:32 did you actually try *reverting* c55fa3cccbc2c672e7f118be8f7484e53a8e9e77 ? Jun 14 09:32:56 where did you add that 'try again later' printk? Jun 14 09:33:03 er, at the comment Jun 14 09:33:04 dog Jun 14 09:33:05 dohg Jun 14 09:33:07 ffs Jun 14 09:33:10 * dwmw2_gone goes to get more coffee Jun 14 09:33:13 g Jun 14 09:37:34 https://github.com/ldir-EDB0/openwrt/commit/96a46c246b085cde9675bcc61feb7cdd73d128b8 is all the extra debug logging. Jun 14 09:38:03 https://github.com/ldir-EDB0/openwrt/commit/befe8f0d01e69230030797fbc9f4360841e566da is our reply to c55fa3cccbc2c672e7f118be8f7484e53a8e9e77 Jun 14 09:40:35 but as you'll note we never see 'printk("KDB: vcc_sendmsg() - entry\n");' so c55fa3cccbc2c672e7f118be8f7484e53a8e9e77 and our fix for it, seems somewhat irrelevant. Jun 14 09:40:56 I would actually *revert* it Jun 14 09:41:02 not attempt to deal with it Jun 14 09:47:46 did you establish *why* pppoatm_send() is returning failure? Jun 14 09:48:06 case e_vc: Jun 14 09:48:06 if (!pppoatm_may_send(pvcc, skb->truesize)) Jun 14 09:48:06 goto nospace; Jun 14 09:48:08 that one? Jun 14 09:49:20 * dwmw2_gone looks at the 'git blame' for pppoatm_may_send() and runs away screaming Jun 14 09:50:02 the logging you see is as of this morning.... we/I've not yet looked in any detail at the logs - you're ahead on that front! Jun 14 09:50:28 ok Jun 14 09:52:02 I woke up to 'yet more printk for you' - built an image, flashed, grabbed logs. I'm up with the parents-in-law and the hh5a is 250 miles away in our hall. Jun 14 09:52:16 Remote debugging :-) Jun 14 09:52:26 heh Jun 14 09:52:41 I really should send you a Solos. We can put that in a PC and do proper git-bisect on the kernel etc. Jun 14 09:52:53 and I have to be a little bit social too Jun 14 09:53:04 *grump* :-) Jun 14 09:53:12 heh Jun 14 09:53:35 do you not have an adsl line? Jun 14 09:55:02 vdsl here now Jun 14 09:55:04 same as you Jun 14 09:55:04 this may all just work on kernel atm anyway... remember the lantiq adsl is an out of tree 'extra'. Jun 14 09:55:19 it might Jun 14 09:55:32 but where you're looking now is not hw-specific and the hw is actually working for llc encaps Jun 14 09:55:51 *if* we can repro on solos in a PC it gets a shitload easier to test with Jun 14 09:56:00 yeah, that's what makes it so flipping confusing! Jun 14 10:11:55 let's add some printk around the case e_llc & e_vc area and see stuff around headroom & truesize Jun 14 10:29:33 ldir: and in pppoatm_may_send to work out what/why it's returning Jun 14 10:47:27 ooooooh - nospace. Jun 14 10:47:43 wonder what llc does.... /me tries Jun 14 11:00:10 hmm, it doesn't do nospace much at all, none in the early stages. Mostly it says 'no headroom' . Right, time to add stuff in pppoatm_may_send as well like wot dwmw2_gone said :-) Jun 14 11:06:09 I wonder if ppp changes have rendered the logic in pppoatm_may_send() obsolete? Jun 14 11:06:52 - spin_unlock_bh(&pch->downl); Jun 14 11:06:53 + spin_unlock(&pch->downl); Jun 14 11:06:57 in 97fcc193f67e584dc6564767c6e186fe1ecd71d2] Jun 14 11:07:19 no, it's asserted that BHs are already blocked Jun 14 11:10:05 is pppoatm_pop() ever getting called? Jun 14 11:10:15 what is atomic_read(&pvcc->inflight) ? Jun 14 11:10:22 you are only ever sending *two* frames correctly, right? Jun 14 11:11:00 in the vc case I think you could be right.... it's a feeling I get too. Jun 14 11:13:21 this is correct isn't it? printk("pppoatm_may_send() pvcc=0x%p size=0x%x\n", pvcc->atmvcc, size); Jun 14 11:14:56 printk("pppoatm_may_send() pvcc=0x%p size=0x%x inflight=%d blocked=%d\n", pvcc->atmvcc, size, 2-atomic_read(&pvcc->inflight), test_bit(BLOCKED, &pvcc->blocked); Jun 14 11:18:39 ok you 'c' much better 'n me :-) Jun 14 11:19:24 it could happen if something is just freeing the skb instead of popping it correctly Jun 14 11:20:46 You could try messing with skb->destructor to catch something freeing it Jun 14 11:20:55 set it when you bump inflight for that skb Jun 14 11:20:58 clear it in pppoatm_pop Jun 14 11:21:22 see previous comment! Jun 14 11:22:46 what is 2-atomic_read(&pvcc->inflight) Jun 14 11:24:40 see the big comment around NONE_INFLIGHT Jun 14 11:25:03 the atomic_t is -2 for no packets queued, -1 for one packet queued, 0 for two packets queued Jun 14 11:25:13 er Jun 14 11:25:24 should be 2+atomic_read() then, right? Jun 14 11:27:33 are you rebuilding whole kernels or just modules? Jun 14 11:27:36 I'll have a look :-) Thanks for your help on this. Sorry I'm not up to a useful standard and need hand holding every step. /me wonders how he gets himself into these positions :-) Jun 14 11:27:58 whole kernels - well, whole images really, which then get flashed. Jun 14 11:32:33 2+atomic_read() feels better and makes sense to me bearing in mind the comment Jun 14 11:35:02 something like http://david.woodhou.se/poptest Jun 14 11:36:10 gaaaah! Wow! ok, that'll be the next build :-) Jun 14 11:36:54 thanks :-) Jun 14 11:39:06 and of course a printk in pppoatm_pop() so we see when it's being called. You added that already, right? Jun 14 11:40:18 of course - /me looks around shifty ;-) Jun 14 11:40:35 jow: flow offload fixes (including the new one from today) backported to 18.06 Jun 14 11:45:32 * ldir doesn't even copy'n'paste so well.. compile error.. ok what did I do. :-) Jun 14 11:55:17 * ldir does a git blame pppoatm.c - anyone ever heard of Linux Torvalds & David Woodhouse? Jun 14 11:55:27 never Jun 14 11:55:35 lol Jun 14 11:55:49 https://twitter.com/dwmw2/status/1007207377509208064 Jun 14 11:56:03 * dwmw2_gone blames PeterZ Jun 14 11:58:52 now it may be me, but that seems a bit harsh :-) /me makes note to never comment code. Jun 14 11:59:33 I actually wonder if it's the DROP_PACKET bit at the end of pppoatm_send() Jun 14 11:59:42 in pppoatm_may_send() we bump the 'inflight' variable Jun 14 12:00:01 we get to (near) the end of pppoatm_may_send() and we call ->vcc->send() Jun 14 12:00:05 3ac108006f (Krzysztof Mazur 2012-11-06 23:17:00 +0100 370) ret = ATM_SKB(skb)->vcc->send(ATM_SKB(skb)->vcc, skb) Jun 14 12:00:05 ^1da177e4c (Linus Torvalds 2005-04-16 15:20:36 -0700 371) ? DROP_PACKET : 1; Jun 14 12:00:40 If that *fails* and we return DROP_PACKET... shouldn't we also deduct one from 'inflight'? Jun 14 12:00:55 morning Jun 14 12:02:26 ^1da177e4c (Linus Torvalds 2005-04-16 15:20:36 -0700 298) * we should really drop the packet, but the generic layer doesn't Jun 14 12:02:26 ^1da177e4c (Linus Torvalds 2005-04-16 15:20:36 -0700 299) * support this yet. We just return 'DROP_PACKET' which we actually define Jun 14 12:02:26 ^1da177e4c (Linus Torvalds 2005-04-16 15:20:36 -0700 300) * as success, just to be clear what we're really doing. Jun 14 12:02:26 ^1da177e4c (Linus Torvalds 2005-04-16 15:20:36 -0700 301) */ Jun 14 12:02:29 ^1da177e4c (Linus Torvalds 2005-04-16 15:20:36 -0700 302) #define DROP_PACKET 1 Jun 14 12:02:30 yeah Jun 14 12:02:32 because that's *really* fucking clear Jun 14 12:03:13 * karlp laughs Jun 14 12:03:27 I think both DROP_PACKET return cases there are fucked Jun 14 12:05:31 without looking/understanding the code - if we dropped it..did it ever actually get into the state of being in flight ? Jun 14 12:06:37 nope Jun 14 12:06:40 but we counted it anyway Jun 14 12:06:46 and thus we won't accept any more Jun 14 12:06:48 ahh, oops Jun 14 12:06:55 that's one thing that might explain this Jun 14 12:08:48 A friend of mine keeps sending me this https://pbs.twimg.com/media/DJyAs4rXoAIqyMb.jpg Jun 14 12:11:59 http://david.woodhou.se/poptest2 should actually build Jun 14 12:12:25 heh Jun 14 12:13:44 thank you - family intervening - will be a while before test - sorry :-) Jun 14 12:35:52 I noticed that a while back the "ndisc" suite of packages ("rdisc6", "ndisc6", etc.) were dropped to oldpackages. Was this just because of lack of maintainership? Jun 14 12:42:02 nbd: can we please get the 11r fixes into 18.06? Jun 14 13:02:15 dwmw2_gone: you mean the 802.11r features added to wpad-mini? Jun 14 13:13:20 with the luci checkbox for 802.11r, my old macbook couldn't log on anymore, it kept thinking the AP was some 802.11x secured AP. Jun 14 13:13:29 on master as of ~week ago Jun 14 13:13:45 (a very old, very unsupported macbook) Jun 14 13:39:07 mkresin: it would be a good start to just get the config fixes in Jun 14 13:39:52 or did that already happen? Jun 14 13:40:12 ah, I think it did Jun 14 13:40:31 so yeah, the wpad-mini fix would also be nice (in master and 18.06) Jun 14 13:45:10 I've set up wireguard on my router and I want to pipe DNS queries through the tunnel between my client and the router. What do I need to set up in order to facilitate this? Jun 14 13:45:56 I want to prevent DNS leaks. Jun 14 14:09:51 dangole: hey, handling calibration data should be handled in a more device specific way ("brcm2708: add RPi3+ wifi calibration data") Jun 14 14:10:06 dan-: you can't blindly install calibration device data for device X for all existing devices Jun 14 14:10:32 dangole: it should probably be some package specific to the RPi3+ Jun 14 14:10:37 dangole: and selected for RPi3+ only Jun 14 14:12:18 dwmw2_gone: adding more than the bare minimum to wpad-mini feels wrong to me Jun 14 14:13:03 mkresin: maybe we need wpad-tiny for -tiny subtargets Jun 14 14:13:33 that said I should look at 11r again Jun 14 14:16:20 the distinction in the help text at least doesn't say "bare minimum" it says, "doesn't include the enterprise stuff that requires an SSL implementation"" Jun 14 14:16:21 stintel: we can just reuse the tiny feature flag Jun 14 14:16:33 you can change the definition again, but that's all -mini is _documented_ to be omitting. Jun 14 14:16:43 select 11r if not tiny Jun 14 14:18:18 jow: interesting Jun 14 14:18:48 that said I'm leaning against 11r, at least on tiny targets Jun 14 14:18:51 sounds like a good compromise between not adding 11r at all and adding it unconditionally Jun 14 14:19:10 in some cases it took weeks to save a few KB here or there, only to spend them for a non-critical feature Jun 14 14:19:36 personally I don't care, I don't use -mini Jun 14 14:19:36 if we restrict 11r to more capable targets then fine with me Jun 14 14:22:38 "select 11r if not tiny" might end in a support hell. default install behaves different depending on target/subtarget/tiny flag Jun 14 14:23:41 yes Jun 14 14:26:04 dwmw2_gone: is your issue the size of -full or just the fact that you have to install one more package to get 11r working? Jun 14 14:27:23 the fact that I have to install one more non-default package to get 11r working Jun 14 14:27:32 if we want to install -full on every dual-radio box by default, that's fine Jun 14 14:27:57 11r should work OOTB on a dual-radio box Jun 14 14:34:59 should work (with user config to turn it on) OOTB or should be enabled and working OOTB? Jun 14 14:35:35 karlp: the first is the current state Jun 14 14:36:11 karlp: well, kind of. requires wpad-full installed Jun 14 14:38:14 you don't need to do anythign more tahn put thetickbox in on both wireless interfaces in luci? Jun 14 14:38:23 never considered using 11r for a single dual radio ap. at least not without working band steering Jun 14 14:38:36 as of a few days ago at least, that made a macbook here unable to connect to the AP at all. Jun 14 14:41:22 dangole: did you see my messages? you were disconnected afterwards Jun 14 14:42:43 rmilecki: no, i missed those. changed location in that fusion festival built-up i'm participating in right now Jun 14 14:43:06 [16:07] dangole: hey, handling calibration data should be handled in a more device specific way ("brcm2708: add RPi3+ wifi calibration data") Jun 14 14:43:07 [16:08] dan-: you can't blindly install calibration device data for device X for all existing devices Jun 14 14:43:09 [16:08] dangole: it should probably be some package specific to the RPi3+ Jun 14 14:43:10 [16:08] dangole: and selected for RPi3+ only Jun 14 14:52:37 mkresin: jow: What about simply installing full wpad on non-tiny targets? It's not a massive burden for for 8M+ devices Jun 14 14:53:50 with which ssl library? :) Jun 14 14:54:12 Unless one is going to be installed by default, the internal one Jun 14 14:55:21 Not too concerned about seeing 11r in wpad-mini, though.. 11w on the other hand needs to be in all variants Jun 14 14:55:35 rmilecki: the whole brcm2708 target seems to be a mess in that regard... Jun 14 14:55:41 when did it get one by default? I thought the current config setupmade you choose either wpad-full-openssl or -woflssl? Jun 14 14:56:15 rmilecki: i was relieved to see that the MAC addresses in target/linux/brcm2708/base-files/lib/firmware/brcm/brcmfmac43430-sdio.txt aren't actually being used Jun 14 14:56:19 on 18.06 and earlier it just explicitly pulls in openssl Jun 14 14:57:50 rmilecki: currently, all arm distros explicitely supporting the RPi3 boards just ship those files in /lib/firmware Jun 14 14:58:12 karlp: Not sure when it got in there but it has an option for internal SSL support, no external library Jun 14 14:59:44 karlp, Monkeh: i converted the hostapd/wpa_supplicant into build variants recently. Jun 14 14:59:59 I think I'll send a patch to enable .11W all round later - It adds iirc ~6.5K to wpad-mini and ~9K to wpa_supplicant-mini - and that's all I need to stop using full wpad on 4M devices.. Jun 14 15:00:09 dangole: iirc it had the internal option before that Jun 14 15:00:41 karlp, Monkey: and as 'internal' was the default choice for wpad and hostapd (full variants), this is now also used for the 'wpa-supplicant', 'wpad' and 'hostapd' packages Jun 14 15:00:59 karlp, Monkey: if you want to use openssl or wolfssl, you got to use those package variants instead Jun 14 15:01:15 the best might be a mix of all. rename wpad-mini to wpad-tiny. Include wpad-tiny for space constrained targets. add 11r to wpad-mini Jun 14 15:01:53 karlp, Monkey: and for various post-WPA2 features (SAE, OWE, HS20/OSEN, ...) you need to use an external library because EC and stuff hasn't been implemented in the built-in internal crypto Jun 14 15:02:19 dangole: It's still hard enough to get people to use WPA2.. :P Jun 14 15:02:21 could still end in a support hell, but at least different named pkg are used Jun 14 15:02:52 mkresin: Long as I still get 11w in -tiny :P Jun 14 15:02:57 yeah, let's have some more hostapd/wpa_supplicant variants Jun 14 15:03:07 dangole: More choice more better! Jun 14 15:03:42 LTO seems to have a respectable impact on size, that may be worth enabling Jun 14 15:06:39 rmilecki: I guess I should partially revert commit c3d3111831 ("brcmfmac43430-firmware: remove package and switch to linux-firmware") and put the calibration txt and blobs into board-specific packages Jun 14 15:07:20 rmilecki: i got no idea which board the existing brcmfmac43430-sdio.txt belongs to though, that RPi3+ was the first of its kind i thouched Jun 14 15:11:36 dangole, For RPi 3B+ it's brcmfmac43455 Jun 14 15:16:57 dangole: i agree brcm2708 is a mess ;) Jun 14 15:17:46 ahhhh - walk into local town away from the keyboard in pretty ok weather - bliss. Jun 14 15:54:32 <_lore_> ldir: regarding acs on mt7603, it is currently not supported Jun 14 15:55:00 _lore_: ahh, that explains why it doesn't work :-) Jun 14 15:55:32 <_lore_> :) Jun 14 15:55:55 behaviour as expected - no fault found. Jun 14 16:12:37 any quick way to extend the 3-min sliding GUI window for network statistics on luci, to say 10 minutes? Jun 14 16:13:32 i have a client dropping wifi with 18.06 and i want to it over a long period of time comparing to another client in parallel Jun 14 16:14:08 basically i need draw a network connection status diagram over time for all clients Jun 14 16:17:27 rmilecki: hope you like it more now after commit 037ef13a16 Jun 14 19:08:24 build #701 of armvirt/64 is complete: Failure [failed dirupload] Build details are at http://phase1.builds.lede-project.org/builders/armvirt%2F64/builds/701 blamelist: Felix Fietkau , Daniel Golle Jun 14 19:13:21 hi. I want to have a virtual package which installs default local settings. is there a way to make sure that it runs after other packages, in particular base-files? Jun 14 19:14:12 and is bobafethotmail ever on this channel, and if so, what's his handle? had a question about some wiki content... Jun 14 19:35:04 hello openwrt ! i got a question, i guess its more router related not openwrt Jun 14 19:35:28 stock firmware, dd-wrt and openwrt on my tp-link n750 wdr4300 have the following behaviour: Jun 14 19:36:02 i can ping like 16k pings to the internet from the router, i cn ping the router with 16k, but i cannot reliably ping from my pc to the internet with even 64byte# Jun 14 19:36:13 wtf is that any ideas Jun 14 20:27:49 is there a way to patch uboot for a given target? Jun 14 20:28:19 (support for that target has been mainlined, but not yet included in a release) Jun 14 20:29:57 azarus: all u-boot builds in package/boot/uboot-* are target specific. (?) Jun 14 20:30:23 azarus: adding another target-specific u-boot build is fine as long as it's not based on a terribly outdated upstream U-Boot Jun 14 20:30:38 dangole: oh, awesome. thanks for lettinge me know! Jun 14 20:32:11 I could just patch the uboot for mvebu, on which it's based... Jun 14 20:33:06 argh, no. the vendor uboot is based off 2013.01 >_< Jun 14 20:34:17 azarus: which target is this all about? Jun 14 20:34:33 dangole: helios4, which is not in OpenWrt yet. Jun 14 20:35:30 dangole: https://kobol.io/helios4 Jun 14 20:35:42 and my beginnings of a port of OpenWrt: https://code.azarus.ch/openwrt/log/?h=helios4 Jun 14 20:35:43 azarus: it should be easy to add this to mvebu and add what ever needed to make uboot-mvebu work Jun 14 20:36:48 Yeah, I'll be able to do it probably Jun 14 20:37:13 uboot seems to be the only thing missing now, already patched the kernel which was easy Jun 14 20:38:23 azarus: marvel u-boot has always been pretty clean (at least back in the days with kirkwood, which was the last time i touched it...) Jun 14 20:39:41 dangole: yay, will have the time to do it tomorrow :) Jun 14 20:45:08 dangole: https://github.com/openwrt/openwrt/commit/6fc8e06078d30e8d36a00d0ecc97ac9cc148fe60 breaks on my end Jun 14 20:46:52 magix: can you provide logs on pastebin? (it does work for me) Jun 14 20:47:18 yeah Jun 14 20:47:30 -Wno-implicit-fallthrough is not being passed Jun 14 20:48:08 I guess TARGET_CFLAGS are not relevant for host builds Jun 14 20:48:09 Hm, how's the easiest way to extract the sources of a package, so one can patch it? Jun 14 20:48:16 I'd like to patch uboot-mvebu Jun 14 20:48:59 dangole: https://gist.github.com/neheb/2c0e413f0cdbf4c7c7f502b5bf25892e Jun 14 20:50:30 ah, answered my own question. Jun 14 20:55:22 magix: should be fixed Jun 14 21:05:07 dangole: my suggestion is to use PKG_ASLR_PIE + backporting the patch fixing this. Jun 14 21:08:08 okay -- this might not be as easy as I thought. patching changes of u-boot-2013.01 to u-boot-2018.03 isn't really trivial... Jun 14 21:09:22 azarus: why don't You take this patch: http://git.denx.de/?p=u-boot.git;a=commitdiff;h=ae28a5f830936ca285ec65d3f5f70fdcef95ab92 Jun 14 21:10:16 tmn505: sometimes I wonder how I don't look in the obvious places... Jun 14 21:10:17 Thank you. Jun 14 21:10:53 azarus: where the U-Boot is stored on this device Jun 14 21:10:58 ? Jun 14 21:11:16 tmn505: can be multiple places. it has boot selector dip switches Jun 14 21:11:24 most people will use SD card booting tough Jun 14 21:11:56 dangole: since you're here, can you ACK this? https://github.com/openwrt/packages/pull/6254#partial-pull-merging Jun 14 21:11:59 doesn't it have nor? Jun 14 21:12:15 tmn505: it does Jun 14 21:12:17 32MB Jun 14 21:13:05 with so much space You can store whole U-Boot with OpenWrt Jun 14 21:13:33 on it Jun 14 21:14:23 sure, how would one flash it tough? boot a debian live image and just flash it over the flash? Jun 14 21:14:39 argh, language is broken today Jun 14 21:15:11 that could be one of the posibilities Jun 14 21:15:43 i wonder what comes preflashed on the SPI NOR Jun 14 21:15:51 not noted anywhere! Jun 14 21:16:16 2nd would be launching U-Boot cmdline and download image from tftp and writing it to flash Jun 14 21:16:30 or, over serial? Jun 14 21:16:49 no, over network Jun 14 21:17:08 tftp is of course over the network, but maybe there's also a way over serial Jun 14 21:17:42 might be. Jun 14 21:18:08 Of course using sdcard would be safer Jun 14 21:18:20 yup, and easier for most people? Jun 14 21:18:49 Yes, for references look at clearfog definitions how to prepare the image Jun 14 21:19:16 mangix: merged the gawk update. thanks for notifying Jun 14 21:19:19 * with u-boot Jun 14 21:19:29 * dangole is off to bed Jun 14 21:19:32 tmn505: Sure, where are they defined? Jun 14 21:19:50 lemme check Jun 14 21:21:17 azarus: it starts here -> https://git.openwrt.org/?p=openwrt/openwrt.git;a=blob;f=target/linux/mvebu/image/cortex-a9.mk;h=6bc6269da6a0b11c4c136d3143e06312346f0097;hb=HEAD#l138 Jun 14 21:22:42 tmn505: seems simple enough Jun 14 21:23:15 tmn505: I only have sdcard for now: https://code.azarus.ch/openwrt/tree/target/linux/mvebu/image/cortex-a9.mk?h=helios4&id=f37bc54b6f9abf816f4da919f3c5eab33b26e9dc#n174 Jun 14 21:23:15 yes it is, there's also one starting at line #125, theyr'e almost same Jun 14 21:24:54 You are missing command sequences to create the image, they are defined at IMAGE/sdcard.img.gz := blah | blah | blah Jun 14 21:25:33 would "IMAGE/sdcard.img.gz := boot-scr | boot-img | sdcard-img | gzip | append-metadata" be okay there? Jun 14 21:25:47 doesn't seem to be anything device specific there Jun 14 21:25:56 yep Jun 14 21:26:00 Cool! Jun 14 21:28:00 to see what most of those commands do look in for example: for boot-scr -> https://git.openwrt.org/?p=openwrt/openwrt.git;a=blob;f=target/linux/mvebu/image/Makefile;h=be01623e86b737ebd29db70b299a0b863d6cc4ad;hb=HEAD#l28 Jun 14 21:28:35 So it mostly just calls mkimage, I presume Jun 14 21:29:35 yes Jun 14 21:32:54 tmn505: okay, fixed the sdcard.img.gz line: https://code.azarus.ch/openwrt/tree/target/linux/mvebu/image/cortex-a9.mk?h=helios4&id=17a6e639a74dd214a3eacbc1b4acb67ff3c3a0bf#n180 Jun 14 21:33:11 and backported helios4 support: https://code.azarus.ch/openwrt/commit/?h=helios4&id=08f402ad8bedca940c7721026304102176d626ec Jun 14 21:34:33 looks good, do You have the device? Jun 14 21:34:51 tmn505: not yet. but I will get it and test it before officially submitting my work :) Jun 14 21:35:57 then good luck, I'll submit my review if You'll post it on ML or Github. Jun 14 21:36:09 tmn505: Yay, thank you :D Jun 14 21:42:29 hey @mangix, actually a board i am testing with has the wifi modules soldered onto the board. but to your knowledge, isn't the mt7621 all pcie, and not pci? wouldn't that mean i should read in "device-type = pcie" Jun 14 21:43:52 biangbiangmian: device-type = "pci" is deprecated and should not be used Jun 14 21:44:15 ah, what's your recommendation for smt wifi? Jun 14 21:44:45 are you trying to add a new device? Jun 14 21:45:06 https://github.com/devicetree-org/devicetree-specification/blob/master/source/devicetree-basics.rst#device_type-deprecated Jun 14 21:45:08 i see, the entire device_type is deprecated Jun 14 21:45:42 i have a mt7621 SoC, and mt7662e + mt7603e wifi modules Jun 14 21:46:33 so i'm just trying to get wifi working Jun 14 21:47:02 what's the error? Jun 14 21:47:35 i'll have that in front of me shortly, but from what i remember it said cant' find sysfs path of radio0 or something Jun 14 21:48:25 that sounds like a userspace issue Jun 14 21:48:33 what does dmesg say? Jun 14 21:48:45 brb in around 10-15 minutes, changing locations Jun 14 21:48:49 it's in my post https://forum.lede-project.org/t/mt7621-pcie-dts-issues-mt76-stopped-working-with-latest-updates/14623/124 Jun 14 21:50:24 dmesg not dts Jun 14 21:53:43 back Jun 14 22:03:51 is the wifi on that SoC usually on pcie 1 and 2, or 0 and 1? Jun 14 22:06:01 i have the board now, so any kind of testing you want me to do, i can do Jun 14 22:08:12 i have a working sample and a not working sample Jun 14 22:08:16 dmesg says Jun 14 22:08:42 bus=0x1, slot = 0x0, irq=0xff Jun 14 22:08:50 bus=0x2, slot = 0x1, irq=0xff Jun 14 22:09:01 and it finds the firmware and enables the module Jun 14 22:10:16 on both? Jun 14 22:11:01 the wifi is found on whereever the manufacturer chose Jun 14 22:15:10 just confirmed this is on the one that's not working Jun 14 22:17:32 but should i have anything in my .dts file regarding pcie? Jun 14 22:17:46 except for status = okay Jun 14 22:18:23 i remember you said i shouldn't have to touch the .dts file Jun 14 22:23:56 so are you adding a new device? Jun 14 22:24:06 hi Jun 14 22:26:26 that new offload feature... how does it decide which connections get offloaded? for example i have ~ 130 connections (established and forwarded) but only 8 of them are offloaded? Jun 14 22:33:04 Heyas everyone. Just wanted to ask if something is possible in general.. I want my router to not accept certain addresses on WAN (as in, if upstream offers it on DHCP request, it will not take it) Jun 14 22:34:24 interesting question. Jun 14 22:35:56 The problem is that when my modem is starting up, if it doesn't have a connection when my router asks for an IP, it will offer an internal IP Jun 14 22:36:33 So then when the connection comes up, it still doesn't work and I have to release the IP manually and reboot the modem Jun 14 22:36:40 DHCP protocol doesn't have anything like a "dhcp deny" Jun 14 22:38:39 If not, I may have to set something up so if it gets that IP, it will release the IP, cycle power on the modem somehow and wait 2 minutes before trying again :P Jun 14 22:38:49 wait, that's a bit weird what you're describing. Jun 14 22:39:07 The modem isn't a true modem, it's a ISP router with "modem mode" Jun 14 22:39:51 so you talk pppoe to your modem? Jun 14 22:40:11 No, it's just ethernet Jun 14 22:41:01 so why does it do dhcp on its own? Jun 14 22:41:13 Because it is a stupid ISP modem/router :P Jun 14 22:41:20 so it's a router. Jun 14 22:42:09 In "modem mode", so it should just pass-through. And it does, providing my router doesn't request an IP at the wrong time Jun 14 22:42:19 I guess it doesn't get to modem mode very quickly.. Jun 14 22:43:22 so it's a router doing dhcp until it's connected, after that it's a passthrough device (and you presumably get dhcp from your ISP)? Jun 14 22:43:38 Yes, that seems to be the case Jun 14 22:43:46 that's several kinds of fucked up. Jun 14 22:43:50 Tell me about it Jun 14 22:44:03 anyway. I assume you're running OpenWrt then? Jun 14 22:44:08 It gets better though :P Jun 14 22:44:33 When they first released it, just browsing the web would trigger its "flood detection".. and cut you off Jun 14 22:44:54 does it have a rotary dial, perchance? Jun 14 22:44:56 Yes and please don't ask what version. It is way out of date and I will be updating it soon >.> Jun 14 22:45:17 I thought support for my router was dropped after the version it is on :P Jun 14 22:45:40 anyway, I can think of a few hacks you could do. Jun 14 22:46:08 Which is why I'm just asking in general, as it might be something I can only do when it's on a recent version :) Jun 14 22:47:29 a simple one would be to restart WAN until it gets an IP address from a public subnet. Jun 14 22:48:03 That doesn't work.. once the modem serves the internal IP, it "remembers" it and will only ever serve that Jun 14 22:48:24 Even when the connection is back up Jun 14 22:48:53 As you say.. several kinds! :P Jun 14 22:49:03 how would you ever expect to get a usable IP through DHCP then? Jun 14 22:49:20 It might not remember it if the router doesn't accept it Jun 14 22:50:29 I was wrong, though. Jun 14 22:50:46 there is DHCPDECLINE in the dhcp protocol. Jun 14 22:53:57 I have no idea how to convince openwrt's dhcp client to issue it for certain subnets, though. Jun 14 22:54:21 Mhm.. it's a super edge-casey thing so I wouldn't be surprised if there isn't something available Jun 14 22:54:34 This is meant to run on machines with restricted hardware after all Jun 14 22:56:34 well, OpenWrt uses busybox udhcpc as dhcp client, if I'm not mistaken. Jun 14 23:00:37 but, ugh, way out of my depth there. I'm not sure if you can tell it (presumably through an option) to reject certain addresses. Jun 14 23:01:11 It's okie, appreciate the discussion nonetheless :) Jun 14 23:01:56 honestly, though, you should kick your ISP into a suitable body part to give you a sane modem. Jun 14 23:02:43 :P Jun 14 23:02:48 what ISP is that anyway, out of sheer curiosity? Jun 14 23:03:16 The first version didn't even come with Modem Mode.. and neither did v2 Jun 14 23:03:18 (if you rather don't say, that's okay.) Jun 14 23:03:20 yes, it's a new device, it's a custom PCB i had built Jun 14 23:03:25 So they don't really pay attention :P Jun 14 23:03:29 Virgin Media Jun 14 23:03:31 there is "-s PROG Run PROG at DHCP events (default /usr/share/udhcpc/default.script)", but I wouldn't ever think about aiming that low Jun 14 23:04:56 pkgadd: Yeah, I saw that but it doesn't say if that can influence the negotiation Jun 14 23:05:28 Reading the default implies it doesn't as it only has handling for successes Jun 14 23:06:52 because that's all it needs to do in all normal or remotely sane cases, but who knows what you could do Jun 14 23:08:22 which part of a dts entry corresponds to how the kernel sees a device? Jun 14 23:08:32 pcie1 { mt76@2,0 Jun 14 23:08:42 the pcie#, is that bus or slot? Jun 14 23:08:56 and the @# is that bus or slot, or something else? Jun 14 23:09:23 [dmesg] bus=0x2, slot = 0x1, irq=0xff Jun 14 23:15:00 Hm Jun 14 23:16:56 The documentation for it doesn't mention any way of influencing the DHCP client through it Jun 14 23:17:35 ADSL/ VDSL= Jun 14 23:17:36 ? Jun 14 23:30:06 Not sure what you mean? Jun 14 23:31:09 what kind of modem, ADSL, VDSL, cable, 3g, lte, v.92bis..? Jun 14 23:31:11 Do you mean what is my internet connection? It's DOCSIS Jun 14 23:31:24 o.k., there I'm out of my expertise Jun 14 23:32:32 I'm not sure how that's relevant? My router isn't talking DOCSIS or anything Jun 14 23:32:59 internet -(DOCSIS)> ISP modem -(Ethernet)> my router Jun 14 23:34:53 biangbiangmian: bus Jun 14 23:35:18 haha, which one? Jun 14 23:35:37 pcie# should be a bus with a single slot Jun 14 23:35:46 at least i think that's how the driver treats it Jun 14 23:36:21 keep in mind the pcie# entry doesn't matter Jun 14 23:36:29 the driver ignores the DTS file in that sense Jun 14 23:38:53 CandyAngel: in case of ADSL/ VDSL, I would have suggested a decent VDSL modem (or actually several different one) capable of running OpenWrt itself for less than a tenner, but I have no overview over docsis stuff Jun 14 23:39:08 Ahh Jun 14 23:41:19 If I remember correctly, and I probably don't as it was ages ago I looked, I think you have to get ISP co-operation to have another DOCSIS modem and I doubt they would let me :P Jun 14 23:45:35 I checked the script running bit in udhcpc's source code and it doesn't do anything with return values, so that's not going to be useful for this Jun 14 23:47:44 Not return values.. exit codes Jun 14 23:49:24 I mean Jun 14 23:49:55 I can totally use it for the nuclear option (giving my router power control over the modem :P) Jun 15 00:34:48 running zbt1326, when wireless is set to 'auto' channel, it can't be brought-up, also, 2.4Ghz performance sucks, tried various tweaking it's best at 2Mbps or so Jun 15 00:34:56 18.06 snapshot that is Jun 15 00:38:44 most of the time it is less than 1Mbps! Jun 15 01:41:29 so, CONFIG_CC_OPTIMIZE_FOR_SIZE is the default now Jun 15 01:41:41 but... is this in the config so it can be toggled? Jun 15 01:42:14 having to override the default optimise for size stuff in lots of places, config option to set everything to performance instead (or at least not -Os or for size) Jun 15 01:50:29 that's for the kernel Jun 15 01:50:42 not userspace Jun 15 01:55:09 build #26 of mediatek/mt7623 is complete: Failure [failed images] Build details are at http://phase1.builds.lede-project.org/builders/mediatek%2Fmt7623/builds/26 blamelist: Daniel Engberg , Daniel Golle Jun 15 02:04:22 dissent1: you ever get mv_cesa working with openssl? Jun 15 02:18:25 afalg is working for me but no interrupts go to the crypto driver Jun 15 02:27:15 mangix: yeah Jun 15 02:27:38 kernel is kinda more important though heh **** ENDING LOGGING AT Fri Jun 15 03:00:10 2018