**** BEGIN LOGGING AT Fri Apr 16 03:00:29 2021 Apr 16 04:16:07  hello everyone!  my router is linksyswrt 1900ac v2 Apr 16 04:16:08   i hava a problem[  I ran into unsolvable troubles when I used wireless relay Apr 16 04:16:08  tx rate drops to an unacceptable level Apr 16 04:16:09  ``` Apr 16 04:16:09  root@OpenWrt# iwinfo wlan1 assoclist Apr 16 04:16:10 EC:38:8F:AF:E0:28 -28 dBm / -70 dBm (SNR 42) 5760 ms ago Apr 16 04:16:10           RX: 162.0 MBit/s, MCS 12, 40MHz 663335 Pkts. Apr 16 04:16:11          TX: 2.0 MBit/s 384751 Pkts. Apr 16 04:16:11         expected throughput: unknown Apr 16 04:16:12   ``` Apr 16 04:16:12  This is my configuration file Apr 16 04:16:17  Log file logread http://fars.ee/YWpc Apr 16 04:16:17  logread   http://fars.ee/YWpc Apr 16 04:16:17 i guess it a wireless driver problem , so who can I contact? Apr 16 04:25:14 deconf_; nobody. marvell laid off and abandoned development of the driver. Apr 16 04:26:25 I feel desperate about it Apr 16 04:26:42 The frequency of the problem is too high Apr 16 04:51:08 deconf: right. I'm just saying that nobody develops the driver anymore. Apr 16 08:54:32 OpenWrt 21.02 and master comes with LuCI bootstrap by default instead of a new theme OpenWrt 2020? Apr 16 09:37:45 Pepe: that's probably the better/ conservative choice for now. there's quite some regression potential involved Apr 16 09:39:12 I would say quite opposite, because for me it seems that everyone these days is using OpenWrt 2020 theme instead of bootstrap, which is shipped by default. Apr 16 09:40:38 I am undecided. But wouldn't mind making 2020 the default Apr 16 09:41:00 arguably it's the better choice because it is mobile friendlier compared to bootstrap Apr 16 09:41:04 but people will complain Apr 16 09:42:03 People will complain about everything. /s Apr 16 09:44:49 the other problem I see is that openwrt2020 is about three times larger than bootstrap Apr 16 09:44:56 43K vs. 13K Apr 16 09:45:58 that could be fixed by dropping the embedded font Apr 16 09:49:41 jow: i do find the traditional theme to be more legible on non-mobile. Apr 16 09:49:55 i gave 2020 a shot but quickly reverted Apr 16 09:50:18 the color scheme might fit the openwrt colours better but it doesn't improve usability if you ask me Apr 16 09:51:52 the color scheme could be tuned Apr 16 09:52:26 what I (from a maintainer pov) dislike about bootstrap that it is an organically grown big ball of whool Apr 16 09:52:31 and the dropdown menus are annoying Apr 16 09:52:40 hard to tweak for mobile Apr 16 09:52:43 ok Apr 16 09:53:15 I do like the bright color schema and button style of bootstrap, and I favor them over 2020 Apr 16 09:53:46 yeah it contributes to usability Apr 16 09:54:29 I could imagine a "bootstrap 2020" theme which would have the bootstrap color schema, button styles, spacing etc. but maybe have a the same left hand side menu Apr 16 09:54:32 I also reverted after trying 2020, wouldn't want to see it as default. Apr 16 10:05:13 guidosarducci: thanks for the link! Apr 16 10:05:48 I've just posted on dnsmasq list. Apr 16 10:07:38 I think putting the 'block list' in a hash/tree structure and shared memory has to be the answer. Apr 16 10:08:02 ldir: oh, I hadn't got around to posting to the list yet. Thought you might try it out as well and see if it was clear. Apr 16 10:08:55 I haven't posted your link, but I have read it, I certainly get where you're going! Apr 16 10:12:05 ldir: ok, that's good feedback. I'll post something tomorrow then. "hash/tree in shared mem" is something I worked out a while ago, but never got a chance to follow up. Now that folks are talking about OOM again I'm getting impatient! :) Apr 16 10:14:59 As I see it there are 3 interlinked problems - large blocklists consuming memory & increasing latency - 1 you can improve memory consumption by optimising the storage of the list, but you still have to store it. 2 You can improve latency by optimising list storage/ordering. 3 tcp requests fork the whole darn thing and so all the structures that could be shared are considered by the OS as private hence potentially requiring 21 * dnsmasq mem to Apr 16 10:14:59 be available. Apr 16 10:15:47 dnsmasq forks to handle requests? Apr 16 10:16:06 ldir: there was some tree/hash stuff a few years back than appeared ready to go: https://lists.thekelleys.org.uk/pipermail/dnsmasq-discuss/2015q1/009257.html and https://lists.thekelleys.org.uk/pipermail/dnsmasq-discuss/2015q1/009278.html. I asked Simon about this but didn't get a response. Apr 16 10:16:16 * ldir hides Apr 16 10:16:34 jow: dnsmasq forks to handle TCP requests Apr 16 10:16:42 disable tcp requests then :) Apr 16 10:16:46 jow: local tcp requests are handled by a forked copy of dnsmasq Apr 16 10:17:31 ldir: that means outgoing DNS requests via TCP made by dnsmasq itselfß Apr 16 10:17:32 ? Apr 16 10:18:58 jow: that's what I'm not too clear on myself - I *think* it's the local clients, but I am woolly on that. Apr 16 10:23:03 ldir: I thought the child also does the request forwarding if not in cache, and that was why they needed a kludge for the child to communicate resolved RRs back to dnsmasq for caching. I might be hazy on things though.... Apr 16 10:45:11 I've been out of it for so long I'm hazy on everything Apr 16 10:50:51 ldir: email sent, now off to bed! I'm sure Simon will have sorted it all out when I wake up... :-) Apr 16 10:51:06 ROFL! Apr 16 10:56:21 ldir: Only half joking really! Simon gets things done...BTW, if you have a sec could you look at a gemini/layerscape/oxnas fix? I noticed those override and disable the BPF syscall, definitely not good for the 21.02 release! https://github.com/openwrt/openwrt/pull/4068 Apr 16 11:27:22 ldir: Merge my 5.10 bump, maybe? ;) Apr 16 11:45:30 rsalvaterra: will look a bit later - have the plumbers in the house at the mo. Leaky toilet ewwwwwwwwww Apr 16 11:46:03 ldir: Ouch, that sucks… Apr 16 11:46:20 No, wait, that actually stinks. The plumbers suck. :P Apr 16 13:12:32 should hostapd and wpa_supplicant be running if all the radios are disabled? Apr 16 13:13:37 russell--: currently it's the way it works afaik, so yes. Apr 16 13:14:26 * russell-- is looking to save RAM on a 32MB device Apr 16 13:20:36 I just removed symlinks to wpad. Apr 16 13:26:43 yeah, i thought of that too, works! Apr 16 14:22:27 as an anecdote only, all of our downstream luci pages look horrible in the 2020 theme, but are fine in the bootstrap theme, and I've tried pretty hard to not do anything fancy in them. that's purely my problem though as a downstream Apr 16 14:22:44 (but very truly, the 2020 one is better on mobile!) Apr 16 16:12:48 karlp: could you share some examples if you find the time? Maybe a static html example. Apr 16 16:13:49 it's almost definitely just my own screwups trying to hack on things and missing some markup I'm presumed to have. Apr 16 16:14:00 mostly it's just wrong fonts and padding all being different. Apr 16 16:15:06 I don't really know hwer epepe gets the "everyone's running 2020 anyway" data from at least :) Apr 16 16:15:40 judging from forum screenshots one could get the impression Apr 16 16:15:52 but it obviously depends on the demographic Apr 16 16:16:25 tinkerers in the various linksys wrt firmware threads are more likely to customize (install themes) than people using stock openwrt looking for advice on how to configure wifi Apr 16 16:17:09 personally I don't have any strong opinion about either theme, ideally I'd love to have a mixture of both Apr 16 16:17:36 the somewhat clean (as in code) from-scratch CSS of 2020 with the visual appearance of bootstrap Apr 16 16:17:40 plus maybe the lhs menu Apr 16 16:18:12 I'm quite ok with just following what's done. Apr 16 16:18:36 and I think 2020 is a better theme to base customizations on Apr 16 16:18:38 we greatly preferred the look of the bootstrap them to the old "openwrt" theme we used to be based on, but it's just the style of the times :) Apr 16 16:18:59 and yeah, our more modern pages have been easier to adjust to things than some of the stuff we did years and years ago Apr 16 16:19:02 bootstrap really needs some cleanup (and better mobile support) Apr 16 16:19:16 karlp: btw, I've been working on templating Apr 16 16:19:36 so soon there'll be no nested E() soup required anymore Apr 16 16:19:46 oh cool. Apr 16 16:20:03 I found that realllllly hard to adjust to, it was too specific to openwrt's usecase. Apr 16 16:20:25 I've not been doign anything recently anyway though, working on new gen hardware right now, not Ui improvements :) Apr 16 16:51:46 stintel: Hey, thanks for getting back. Some models don't have the 1Gib NAND, I think there's a particular revision where they abandon it + the switch on the back Apr 16 16:52:02 GiB* Apr 16 16:52:20 It's a bit of a pain because you cannot actually fit the kernel onto 8MiB :\ Apr 16 16:52:25 as far as I can tell anyhow Apr 16 16:52:47 U-boot takes up a significant chunk Apr 16 16:53:23 Yeah I did a binwalk -E of the initramfs I had. It's chunky. Not sure if it's because of extra packages or because of libraries necessary for OpenWrt Apr 16 16:53:24 jow: just had a bit of a run around, I've got less problems than I rememebered from last time I did it. Apr 16 16:53:32 Ultimately the best way to boot it is probably via PCIe Apr 16 16:54:24 it's little things like not rendering as a button anymore, it's nothing major. Apr 16 16:54:32 You should try that instead of anything else -- seriously. You just need to blacklist the liquidio driver and use oct-pci-boot from https://github.com/Hurricos/OCTEON-im8724-SDK Apr 16 16:55:17 Or rather, I think oct-pci-boot boots u-boot for you. Ping me back if you want me to upload the compiled u-boot binary somewhere Apr 16 16:57:22 I've got some radio buttons that don't appear anymore either, but it's very likelyu just shoddy html on our end anyway, not anything I could blame openwrt2020 theme for itself :) Apr 16 16:58:02 karlp: thanks for checking. The compat is indeed a known issue Apr 16 16:58:29 that's definitely the most common thing I've noticed. Apr 16 16:58:34 karlp: adding either a class "btn" or using