**** BEGIN LOGGING AT Mon Apr 06 02:59:58 2020 **** ENDING LOGGING AT Mon Apr 06 05:10:20 2020 **** BEGIN LOGGING AT Mon Apr 06 05:12:21 2020 Apr 06 06:25:25 blogic: wow, you did it! hwnat! congratulations :) Apr 06 07:03:23 rmilecki, on what target? Apr 06 07:04:00 nitroshift: mediatek Apr 06 07:04:11 gch981213, thanks Apr 06 07:22:59 I have a Zyxel 2812. Running it on 19.07 and previously also running from trunk. However since a week (after introducing 5.4 kernel) the router is bricked. After reboot I have no ethernet ports, only localhost. Any tips? Apr 06 07:23:50 o, I have used diffconf for upgrading the config file Apr 06 07:26:06 blogic: Thanks for the hwnat! Can it also be used on mt7621? Apr 06 07:28:29 rmilecki: and on its way upstrema Apr 06 07:28:40 Rene__: i will rebase it on ramips when its bumped to v5.4 Apr 06 07:28:43 blogic: like patches sent? Apr 06 07:29:00 no but grabbed a job to send it upstream Apr 06 07:29:11 rmilecki: problem is that flow offload is part of the ethernet driver Apr 06 07:29:23 but with mt7622/9 we can also do wifi offload Apr 06 07:29:40 so we need an extra pdev to desacribe the offload engine Apr 06 07:29:48 and then the drivers can check if they share this Apr 06 07:29:56 oh, fun Apr 06 07:30:10 right now the hw need to check iof the in/egress device both are on the same hwnat Apr 06 07:30:29 gets more complicated when we do qca80 offload where it is only offloading inside DSA Apr 06 07:30:40 but I am on it ;) Apr 06 07:30:53 i am able to shift wirespeed with 256byte frames Apr 06 07:31:00 at 0% cpu load Apr 06 07:31:13 at 128byte frames it drops to ~780mbit Apr 06 07:32:19 blogic: openwrt/master is bumped to v5.4 or are you waiting that it is in v19? Apr 06 07:32:37 Rene__: ramips is on v5.4 ? Apr 06 07:32:44 what is v19 ? Apr 06 07:33:03 v19 = v19.0xx openwrt branch Apr 06 07:33:08 blogic: I've pushed the patches for mt7621 last Saturday. Apr 06 07:33:23 gch981213: ah ok, let me bump the driver in that case Apr 06 07:33:30 is it on DSA ? Apr 06 07:33:35 blogic: Yes. Apr 06 07:34:18 ok, let me go looking for a mt7621 unit Apr 06 07:35:40 image for devices with w25q256 flash can't reboot atm. I'm working on it. Apr 06 07:35:41 mt7621 has also a trick to wifi offload. What I understand is that they just push the wifi packed to packet engine. Apr 06 07:42:03 gch981213: I wasnt unaware about that upstream revert mail, so from my side it's ok as Robert Marko is aware about situation Apr 06 08:05:18 gch981213: is it possable to add SFP port for ER-X-SFP? DENQ already ported some of the patches. But they were not included any more in the last series. Apr 06 08:06:08 This is my branch that I use to test my device https://github.com/vDorst/openwrt/commits/ramips-5.4 Apr 06 08:08:20 I am also happy that NAND issue is resolved and that Weijie sended patches upstream. Apr 06 08:16:58 Rene__: You could send patches/create pull requests for that SFP support. Apr 06 08:18:07 Rene__: Unfortunately Weijie made the difference between mt7621 and other mtk nand controllers too clear, giving upstream a perfect reason to reject it. :P Apr 06 08:24:35 gch981213: how difficult is upstream's suggestion? Apr 06 08:25:34 mangix: Impossible for us and almost impossible for Weijie. Apr 06 08:26:32 Existing mtk_nand driver uses legacy nand APIs and I doubt if upstream will accept this kind of major rework without also replacing legacy APIs. Apr 06 08:27:35 Rewriting that driver requires testing on mt2701 and mt7622. Apr 06 08:29:08 mt2701 is MTK's android chip. I believe we can't obtain a board allowing custom firmware on it. Apr 06 08:33:52 Why does rewriting the driver _require_ testing on an unobtainable chip? Apr 06 08:34:23 gch981213: Yes I can. Apr 06 08:37:15 Why can't your rewrite the driver dropping support for that nasty mt2701 and then it's up to MTK to bring it back if they care. Apr 06 08:38:45 If the vendor actively prevents community maintenance of their driver then it's a fair game for the community to break the support while reworking it. Apr 06 08:40:32 PaulFertser: Oh. I'd say that's fair enough :) Apr 06 08:42:13 hahaha I love that suggestion Apr 06 08:42:42 PaulFertser: lol Apr 06 08:42:54 PaulFertser: you liturally have no clue what you are talking about in regards to mtk Apr 06 08:43:17 utter pile of boolocks, sorry to be so direct Apr 06 08:43:28 gch981213: was it no in car entertainment ? Apr 06 08:43:35 .... which is probably android based Apr 06 08:44:11 blogic: I only know that's not a router chip so I guessed it's for some android devices. Apr 06 08:44:30 blogic: so why should be they be taken into account when reworking the drivers? If they do not allow testing on some SoC, so be it, drop the support, why not? Apr 06 08:44:31 i think it was set top box or in car Apr 06 08:44:49 PaulFertser: do you know the /ignore feature ? Apr 06 08:45:11 blogic: my IRC client has one, yes, why? Apr 06 08:46:43 When I rewrote MTK's crappy spi nor driver using spi-mem, they expressed their concern about their product maintenance via private email. Apr 06 08:47:00 who did that ? Apr 06 08:47:18 i consider mtk to be the community friendliest company Apr 06 08:47:27 they even release datasheets Apr 06 08:47:50 I'd also like to avoid this additional test pressure for them but upstream insist that I remove the old one. Apr 06 08:48:51 blogic: They said it's okay if it's a decision from upstream maintainers :D Apr 06 08:49:32 and they do give me the controller datasheet Apr 06 08:50:34 gch981213: so they just paricipated in the public process raising a concern and agreeing that it is not their call Apr 06 08:50:39 and provided the required docs Apr 06 08:50:52 sounds pretty legit to me Apr 06 08:51:12 gch981213: Should we push the CPU port learing hack? Apr 06 08:51:32 dengqf6: is that related to the mail regarding DSA learning ? Apr 06 08:51:44 blogic: yup Apr 06 08:52:05 Andrew says try enabling MAC learning for CPU port Apr 06 08:52:05 blogic: I'm not saying mediatek is all-around bad. But if they do not allow arbitrary code execution on a given SoC this sounds just nasty. And it shouldn't be something holding back progress on proper upstream support for their other chips. Or is my logic flawed? Apr 06 08:52:14 yeah i read that Apr 06 08:52:31 right now you are hitting the fdb ageing timeout Apr 06 08:52:48 blogic: I've tested that on 2 MT7621 routers Apr 06 08:53:33 PaulFertser: That's just my assumption. There's no android devices on the market without implementing a signature checking mechanism. Apr 06 08:54:15 gch981213: without end user controlling the root of trust it's just hostile towards the users I'd say. Apr 06 08:55:02 PaulFertser: That's not MTK's fault. It's device vendor's decision on whether/how to use it. Apr 06 08:55:27 gch981213: MTK can include some GPLv3 component to prohibit tivoization if they care. Apr 06 08:55:55 gch981213: also MTK can be just selling devboards Apr 06 08:58:04 PaulFertser: GPL doesn't solve anything here. It only enforces code releasing and has nothing to do with code signature verifying. Apr 06 08:58:26 gch981213: GPLv2 doesn't, but GPLv3 does Apr 06 08:58:40 Is jow lurking here? Apr 06 08:58:55 So MTK can release their bootloader under GPLv3 to make vendors be friendly to their users. Apr 06 09:01:37 PaulFertser: this is wishful thinking. This matter isn't in MTK's hands. Apr 06 09:02:34 f00b4r0: so why can't MTK release bootloader and other useful code under GPLv3? Apr 06 09:03:10 speculating but the most likely answer is: it would hurt their _income generating_ business. Apr 06 09:03:13 I'd say no major vendor will be interested in their product if they do so. Anyway these are just minor problems. Nobody other than themselves is interested in rewriting that driver properly. Apr 06 09:03:20 ^ Apr 06 09:03:40 blogic: does your qca8k hwnat driver work? Apr 06 09:07:57 blogic: Saw your wrc2533gent commit. Do we really want to follow MTK using jffs2 on nand flash? Apr 06 09:09:32 I know this caused some serious issues for a vendor. 20% routers from their production line got corrupted jffs2. Apr 06 09:10:34 why did they choose to do so instead of ubi? Apr 06 09:10:43 gch981213: no Apr 06 09:10:45 and they are switching to ubi and reflashing all their routers produced. Apr 06 09:10:54 gch981213: but it requires a uboot update Apr 06 09:11:12 gch981213: uboot uses bmt Apr 06 09:11:15 mkresin: ping Apr 06 09:11:27 so replacing on flash with ubi requires a uboot that can talk ubi Apr 06 09:11:35 blogic: It's still possible to use a separated kernel+ubi approach Apr 06 09:11:51 gch981213: if the kernel is inside a bmt region Apr 06 09:11:58 and then sysupgrade will require bmt support Apr 06 09:12:22 arrrggghhh that unnecessarily complicated bmt stuff.. Apr 06 09:12:33 gch981213: and cutting out hw and telling users to not use openwrt because we dont like the technology is no a good road to take Apr 06 09:12:41 gch981213: yes Apr 06 09:12:55 but what can we do, at least we made it work in a clean way without the drivers Apr 06 09:13:05 gch981213: ideally a staged uboot will do eventually Apr 06 09:13:18 so preloader->uboot->uboot-owrt Apr 06 09:13:28 until then we are stuck with bmt Apr 06 09:14:05 gch981213: i already have the uboot sources however Apr 06 09:14:07 ;) Apr 06 09:14:14 ok Apr 06 09:24:27 blogic: good morning Apr 06 09:26:40 aparcar[m]: hi Apr 06 09:33:03 jow: do you have an opinion on my CDN suggestion? Apr 06 09:36:15 blogic: do you get the DM? Apr 06 10:21:04 * mangix enjoys the progress on mt7621 despite having no hardware Apr 06 10:27:40 someone may suggest me a place where I can finda copy of xrx300_phy22f_a21.bin Apr 06 10:28:17 of course - no copy in vendor firmware, no copy in linux-firmware apparently; I can find only a patch by Hauke adding it :D Apr 06 10:28:17 aparcar[m]: yeah makes sense, but lets use subbdomains Apr 06 10:28:27 aparcar[m]: cdn.sources.openwrt.org Apr 06 10:28:38 aparcar[m]: and eventually cdn.downloads.openwrt.org Apr 06 10:28:42 let me know once it's set up Apr 06 10:28:59 then I can add a https cert Apr 06 10:29:12 alternatively sources.cdn.openwrt.org and downloads.cdn.openwrt.org Apr 06 10:29:22 not sure what is better Apr 06 10:29:32 cdn.sources.openwrt.org please Apr 06 10:30:04 we already got cdn.openwrt.org so that's may break things Apr 06 10:40:11 no ideas? Apr 06 10:41:47 +1 for sources.cdn.openwrt.org Apr 06 10:42:31 but it's just a bikeshedding, doesn't matter Apr 06 10:43:43 Hauke: may you point me out on where I may find xrx300_phy22f_a21.bin ;thanks. Sorry for the noise. Apr 06 10:45:16 aparcar[m]: and what needs to point where now? Apr 06 10:46:26 I need to do ynezz a favor as he reviewed my json patch. please point sources.cdn.openwrt.org to sourcescdn-1212d.kxcdn.com Apr 06 10:46:57 aparcar[m]: no, you've promised the signify :p Apr 06 10:47:38 ynezz: ugh that's difficult because libbsd is big Apr 06 10:49:31 well if we just need the signify, then I guess the static compilation might be a way to go? Apr 06 10:49:48 it probably just needs a one or two functions out of libbsd Apr 06 10:49:58 I assume, that signify shouldnt be a huge libbsd user Apr 06 10:50:22 did not check that Apr 06 10:50:26 building libbsd with -ffunction-sections and linking it statically with -Wl,gc-sections should trim it down considerable Apr 06 10:50:39 dangole: wonna join in the libbsd fun? Apr 06 10:51:02 ynezz: of and I got some kernel patches if you like https://github.com/aparcar/linux Apr 06 10:51:20 then please send them upstream :) Apr 06 10:51:30 never did that Apr 06 10:52:00 just git send to torvalds@linux.org? Apr 06 10:52:27 jow: please ping me if you change the dns Apr 06 10:52:58 aparcar[m]: https://www.kernel.org/doc/html/latest/process/submitting-patches.html Apr 06 10:53:04 aparcar[m]: done Apr 06 10:53:11 and we may should add directly the downloads.cdn.org thing Apr 06 10:53:43 sure, where should it point to? Apr 06 10:53:57 to the same address cdn.openwrt.org is pointing to? Apr 06 10:54:58 currently it is cdn-1212d.kxcdn.com but I can rename it to something else so it matches the new naming system Apr 06 10:55:25 as we may have a cdn for more stuff eventually? Apr 06 10:56:12 then lets rename. the cdn.openwrt.org cname has a ttl of 120 seconds Apr 06 10:56:16 so it should quickly converge Apr 06 10:58:34 uhm so: Apr 06 10:58:35 downloads.cdn.openwrt.org -> downloadscdn-1212d.kxcdn.com Apr 06 10:58:35 sources.cdn.openwrt.org -> sourcescdn-1212d.kxcdn.com Apr 06 10:59:07 okay and cdn.openwrt.org to downloadscdn-1212d.kxcdn.com as well? Apr 06 10:59:23 but then forwarding cdn.openwrt.org to downloads.cdn.openwrt.org does not work - right? Apr 06 10:59:55 *.cdn.openwrt.org and cdn.openwrt.org can point to different places Apr 06 11:00:08 yey Apr 06 11:00:12 I can point cdn.openwrt.org to a vhost on downloads.openwrt.org and redirect from there if KeyCDN does not offer redirect functionality Apr 06 11:00:30 (or aliases) Apr 06 11:00:31 ynezz: thanks for the patch advise, I'll read it tomorrow with my first coffee Apr 06 11:01:02 they have aliases Apr 06 11:01:23 well then I just point downloads.cdn.openwrt.org *and* cdn.openwrt.org to downloadscdn-1212d.kxcdn.com Apr 06 11:02:04 cdn-1212d.kxcdn.com can then be removed Apr 06 11:02:15 ack Apr 06 11:02:38 okay, all don Apr 06 11:02:40 e Apr 06 11:05:04 thanks Apr 06 11:12:42 jow: can you please add the JSON buildbot step as well? Apr 06 11:12:49 maybe remove the haltonfailure Apr 06 11:13:02 sigh... yet another JSON change? Apr 06 11:13:19 will do but likely not today Apr 06 11:13:24 jow: the fun never ends Apr 06 11:23:34 * ldir all persons found to be having fun will be shot ;-) Apr 06 11:25:11 jow: Validation failed. Verify the CNAME in your DNS settings. Apr 06 11:25:24 is is a regular cname for all of them Apr 06 11:27:28 https://pastebin.com/yacSmFca Apr 06 11:34:45 thanks Apr 06 11:34:54 I guess keycdn uses a slower dns... Apr 06 11:35:16 Zone Alias successfully added. It's globally active in approx. 10 minutes. Apr 06 11:35:18 now it works Apr 06 11:35:52 SSL setup needs to be done yet Apr 06 11:37:43 I'm on it Apr 06 11:37:51 problem is only one letsencrypt cert is allowed Apr 06 11:37:57 so no more cdn.openwrt.org Apr 06 11:38:17 sorry for the extra loop Apr 06 11:44:41 okay now https works too Apr 06 11:50:55 great. thank you Apr 06 11:51:14 I suggest to let it settle for one or two days, then follow up with downloads.pl & friends Apr 06 11:53:54 jow: ugh sorry I just send a patch, did not check in here before sending Apr 06 11:58:39 aparcar[m]: we don't have a "fancy" directory listing for sources, so no keycdn sponsorship footer either. Is this going to be a problem? Apr 06 11:59:12 if so we maybe should add a dir list in a similar style to downloads.openwrt.org Apr 06 11:59:49 I doubt it as it's nothing anyone would visit via a browser anyway. I'll let them know to be sure Apr 06 12:05:22 I'm out for today, bye and thanks for the quick setup jow Apr 06 12:05:34 ynezz: signify incoming soonish Apr 06 12:22:53 Ok, in openwrt git tree I can find xrx200_phy22f_a22.bin but the driver requests xrx300_phy22f_a21.bin ... Apr 06 12:28:35 try symlinking, it might be autogenerated based on a probed revision code. Apr 06 12:31:59 karlp: I tried, but device got stuck ... mhm... and I see the filename is hardcoded in the module... Apr 06 12:33:35 https://github.com/ianmacd/d2s/blob/master/drivers/soc/lantiq/gphy.c Apr 06 13:57:13 build #312 of layerscape/armv8_64b is complete: Success [build successful] Build details are at http://buildbot.openwrt.org/master/images/builders/layerscape%2Farmv8_64b/builds/312 Apr 06 14:12:28 karlp: ping Apr 06 14:33:18 mrkiko: iirc openwrt doesn't distribute it for license reasons Apr 06 14:33:52 i did find it in a vendor tarball before, but i no longer remember which Apr 06 14:40:14 mrkiko: this file should contain the PHY FW: https://gitlab.com/gplmirror/telekom-speedport-smart-3/-/blob/master/ugw_sw/dl/ltq_fw_PHY_IP-2.2.tar.gz Apr 06 14:40:40 ldir: whats up? Apr 06 14:41:15 You use umdns I think? Apr 06 14:41:40 yeah? Apr 06 14:42:07 I just wanted to check that the last bumps didn't actually break it :-) Apr 06 14:42:19 havne't checked master recently sorry :) Apr 06 14:42:36 Tapper said it did not compile Apr 06 14:42:44 if ubus call umdns browse or list, or whatever the ubus method is "works" Apr 06 14:42:58 and you can use avahi-browse o see the owrt device, it's "working" Apr 06 14:43:18 DonkeyHotei: that what did not compile? Apr 06 14:43:26 umdns Apr 06 14:43:46 karlp: on that basis, it's fine :-) Apr 06 14:44:46 DonkeyHotei: fake news Apr 06 14:45:25 i just compiled a new build off master like an hour ago Apr 06 14:45:49 (no idea if umdns is enabled by default though) Apr 06 14:47:49 Borromini: "ubus call umdns browse" shoujld "not crash" for starters Apr 06 14:47:53 (if it's installed) Apr 06 14:49:35 avahi-publish -s blah _blah._tcp 9000 in another window should get you some service to browse Apr 06 14:53:08 dns-sd -B _ssh . sees the umdns advertised ssh service on my router from my macbook. Apr 06 14:53:28 then it "works" still :) Apr 06 14:53:35 yarp Apr 06 14:53:46 I don't have fine grained tests on more than that. Apr 06 14:54:21 well, ubus call umdns browse is a second test, those are differet portions Apr 06 15:03:16 * ldir ha hah! Hilariously have managed to get Debian 10.3 on my Asus eeepc running - 768MB ram, 1GHz AMD clockwork toy of processor. Thing that partially saves it - 128GB SSD Apr 06 15:54:49 Hauke: have you heard the news? Intel's Home Gateway division (formerly Lantiq) is being sold to MaxLinear Apr 06 16:34:37 xdarklight: yes I am aware of this Apr 06 16:37:10 ldir: I'm using Toshiba ac100, you can compare the specs :) And 32 GB slow eMMC soldered instead of proper storage. Apr 06 17:18:29 xdarklight: for the peanut price of 150m Apr 06 17:18:36 which is a lot for what they get Apr 06 17:18:40 ;) Apr 06 17:19:06 xdarklight: ppl in the valley are actually shocked about the rock bottom price Apr 06 17:19:26 lots of rumours floating about Apr 06 17:19:45 my personal guess is that they will abandon 50% of the silicon Apr 06 17:20:04 and bought it solely for the current supply chain and lets say gpon IP Apr 06 17:20:25 chances are that the downward spiral will accelerate Apr 06 18:15:17 Making a 6in4 tunnel in Luci gets stuck. Apr 06 18:15:38 snapshot x64-squash-combined-efi Apr 06 18:19:08 Appears any interface generation gets stuck Apr 06 18:24:20 some kind of segfault in rpcd Apr 06 18:31:23 but thats not the only problem Apr 06 18:32:36 no changes to rpcd in months Apr 06 18:32:50 so I guess a recent gcc bump or whatever causes it Apr 06 18:34:02 rpcd crash happend upon installing wireguard Apr 06 18:34:23 via luci? Apr 06 18:34:38 no, via command line Apr 06 18:34:43 let me repro Apr 06 18:37:30 wireguard or luci-app-wireguard? Apr 06 18:37:50 jow: curious to know what openwrt's kernel release schedule is Apr 06 18:37:58 by when do you suspect there will be no more < 5.6? Apr 06 18:38:06 Does that take years because of old device support Apr 06 18:38:11 Or is openwrt pretty fast moving Apr 06 18:38:31 zx2c4: iirc the other devs try to follow lts releases Apr 06 18:39:31 can't repro Apr 06 18:42:00 well, we typically have three kernel versions in parallel Apr 06 18:42:26 I've killed the last 4.9 targets just 1-2 months before 5.4 was added Apr 06 18:43:13 and about the same will apply to stable branches; master + two stable branches will also end up in three different kernel versions Apr 06 18:43:21 so, you can extrapolate Apr 06 18:43:33 jow: ahh LTS, gotcha Apr 06 18:55:30 opkg update uses more than 67 mb free and 45MB available (128 MB ram total) Apr 06 18:55:43 JMV2000: known issue, there's an open ticket Apr 06 18:56:28 I once spent a month profiling it, reducint its memory usage by 60%, fine tuning internal data structures Apr 06 18:56:46 tnks for that. Apr 06 18:56:58 but then something or someone drove a bulldozer over it and now its behaving as it is, didn't have time to investigate Apr 06 18:57:11 my guess it has to do with the vfork stuff Apr 06 18:57:29 blogic: the price is shocking. it means compared with the price that Intel paid for Lantiq they "lost" 100m. and I don't see why that is either Apr 06 19:44:39 xdarklight: intel expects X$ growth Apr 06 19:45:03 anything below that margin is considered suboptimal Apr 06 19:45:18 its imperialistic thinko bollocks Apr 06 19:45:48 its better to loose 100m than not gain 10 / quarter Apr 06 19:46:11 these are slip head detached from reality Apr 06 19:46:35 and all this over a trashcann of shitty indian code called intel chd Apr 06 19:47:13 i mean lets face it, what was previosly called call lantiq is now a traschcan of shit code from india. Apr 06 19:47:36 honestly paying 150m is still well ott Apr 06 19:48:03 the gpon part cant be worth more than half of that Apr 06 19:48:37 if i had the cash i'd pay max 50m for that pile Apr 06 19:49:15 $corp madness Apr 06 19:49:55 human kinds biggest problem is the misundertanding of the exponential function Apr 06 19:50:43 :P Apr 06 19:51:01 Borromini: :P Apr 06 19:51:44 its all monopoly money to them i think Apr 06 19:51:56 detaxhed from reality Apr 06 19:52:05 it is Apr 06 19:52:15 it really is Apr 06 19:52:17 it's numbers, accounting and 'fiscal optimisation' Apr 06 19:52:28 excel tabels Apr 06 19:53:05 a friend of my brother's works at nestlé. all he does is sloshing humoungous amounts of money around to profit off exchange rates and whatnot Apr 06 19:53:06 s/el/le Apr 06 19:53:53 nestlé, he wont be going to heaven Apr 06 19:54:09 -> bed, long day Apr 06 19:54:12 :D Apr 06 19:54:15 who will in the end. Apr 06 19:54:16 goodnight Apr 06 20:11:23 ditto **** ENDING LOGGING AT Tue Apr 07 02:59:57 2020