Wizzup | freemangordon: | 02:08 |
---|---|---|
Wizzup | Unpacking leste-config-common (1.88-1+m7.1) over (1.86-1+m7) ... | 02:08 |
Wizzup | Setting up leste-config-common (1.88-1+m7.1) ... | 02:08 |
Wizzup | Removing 'diversion of /etc/profile.d/xdg-runtime.sh to /etc/profile.d/xdg-runtime.sh.leste-orig by leste-config-common' | 02:08 |
Wizzup | Removing obsolete conffile /etc/profile.d/xdg-runtime.sh.leste ... | 02:08 |
Wizzup | seems like it does? or? | 02:08 |
freemangordon | Wizzup: yes, that's fine | 08:32 |
Wizzup | freemangordon: what I mean is that it does seem to remove the etc file | 11:05 |
Wizzup | or is the 'obsolete conffile' your echo? | 11:05 |
* Wizzup checks | 11:05 | |
Wizzup | ah, maintscript? never heard of it before | 11:07 |
buZz | ooooo, jabber works again \o/ | 12:26 |
buZz | thanks! | 12:27 |
buZz | so some people overclocked SPI interfaces to cards here? what improvement in speeds did that give? | 12:47 |
Wizzup | tmlind: I made a ticket for your find fwiw https://github.com/maemo-leste/bugtracker/issues/704 | 12:50 |
uvos__ | buZz: spi??? | 13:04 |
buZz | uvos__: the interface where SD is accessed over | 13:10 |
uvos__ | lol | 13:10 |
uvos__ | no | 13:10 |
uvos__ | i mean you can access sdcards over spi | 13:11 |
uvos__ | but no high performance device dose that | 13:11 |
uvos__ | that mode is 2.5mb/s max | 13:11 |
uvos__ | meanwhile | 13:11 |
uvos__ | yes overclocking the bus gets you 40 ish mb/s read and write | 13:11 |
uvos__ | on a fast sdcard | 13:11 |
uvos__ | this is noticable when lanuching stuff | 13:12 |
buZz | with just plain SPI mode? | 13:12 |
uvos__ | no | 13:12 |
uvos__ | we dont use spi mode | 13:12 |
uvos__ | forget about that | 13:12 |
uvos__ | its only used by uc | 13:12 |
uvos__ | we use 4bit wide path | 13:12 |
buZz | ah alright | 13:14 |
buZz | can we overclock -that- ? :P | 13:14 |
uvos__ | yes | 13:15 |
buZz | what speed benefits does it show? | 13:15 |
uvos__ | 40ish mb/s | 13:15 |
uvos__ | on a good sdcard | 13:15 |
buZz | vs? | 13:16 |
uvos__ | 20 | 13:16 |
buZz | nice, and you're describing d4? | 13:16 |
uvos__ | n900 should be the same | 13:16 |
uvos__ | same mmc controller | 13:16 |
buZz | :) nice, how can i enable that OC? | 13:17 |
uvos__ | difference isten that great in real usage however, launching sutff ist faster | 13:17 |
uvos__ | but no where near 2x | 13:17 |
uvos__ | probubly cpu is limit | 13:17 |
uvos__ | buZz: you have to hack the dirver atm | 13:17 |
buZz | aw :( | 13:17 |
uvos__ | theres no easy way | 13:17 |
uvos__ | i have been wanting to add as a sysfs option to force a freqency | 13:18 |
uvos__ | but too mutch to do | 13:18 |
buZz | i guess it improves throughput but not accesstimes? | 13:22 |
Wizzup | probably also improves access time a bit under load | 13:24 |
tmlind | Wizzup: thanks i should have filed the bug myself.. | 16:32 |
Wizzup | np, it's good info | 16:36 |
Wizzup | I just don't have an immediate fix right away, but had a note to self not to forget about the find | 16:37 |
Wizzup | I did notice that on wifi the battery does drain a lot faster | 16:37 |
Wizzup | so if this is what causes it... | 16:37 |
Wizzup | oh wait | 16:39 |
Wizzup | I wonder if we can do this with sysctl | 16:51 |
Wizzup | maybe net.ipv4.neigh.default.interval_probe_time_ms | 16:53 |
Wizzup | (and net.ipv6.neigh.default.interval_probe_time_ms) | 16:53 |
Wizzup | tmlind: putting this in /etc/sysctl.d/no_neigh.conf: | 17:45 |
Wizzup | net.ipv4.neigh.default.interval_probe_time_ms = -1 | 17:45 |
Wizzup | net.ipv6.neigh.default.interval_probe_time_ms = -1 | 17:45 |
Wizzup | makes them all -1 (2147483628) | 17:45 |
Wizzup | freemangordon: shall I add this to leste-config ? | 17:52 |
tmlind | Wizzup: no idea if that makes connecting to a network slower or something.. have not seen any issues here | 17:55 |
tmlind | i guess it's like arp related stuff, anybody know? | 17:55 |
bencoh | neigh/default/interval_probe_time_ms - INTEGER | 17:56 |
bencoh | The probe interval for neighbor entries with NTF_MANAGED flag, the min value is 1. | 17:56 |
bencoh | still trying to understand what this actually means | 17:57 |
Wizzup | I think it might be for p2p wifi stuff or something? | 17:57 |
tmlind | no idea if -1 makes any sense, value of 0 did not work via sysfs | 17:57 |
Wizzup | https://man7.org/linux/man-pages/man8/ip-neighbour.8.html | 17:58 |
bencoh | https://lore.kernel.org/netdev/20220629084832.2842973-3-wangyuweihx@gmail.com/ looks like it's new btw (?) | 17:58 |
Wizzup | maybe we just increase it to like 300s or something | 17:59 |
bencoh | do we even have managed entries in a normal usecase (ie. without issueing ip neigh replace commands)? | 18:01 |
* tmlind on d4 right now and not into browsing into the www currently | 18:01 | |
bencoh | (stable's ip command has no support for that thing apparently btw) | 18:03 |
tmlind | maybe neigh is some arp like protocol? | 18:04 |
bencoh | we do have ip neigh (it's the replacement to the arp command indeed, meant to be used with ipv4 arp and ipv6 neigh), just no support for the "managed" flag | 18:05 |
tmlind | ok | 18:06 |
Wizzup | bencoh: ip neighbour did print some stuff on my laptop | 18:06 |
bencoh | Wizzup: with a managed flag? | 18:06 |
Wizzup | let me check | 18:06 |
Wizzup | not sure how to check | 18:07 |
bencoh | (it prints the arp/neighbour tables, basically) | 18:07 |
Wizzup | yeah | 18:07 |
Wizzup | don't know how to see if it is managed | 18:07 |
bencoh | me neither, but I don't think this version knows how to anyway | 18:07 |
bencoh | https://lore.kernel.org/netdev/20211011121238.25542-1-daniel@iogearbox.net/ | 18:09 |
tmlind | try arp -a or whatever the command is | 18:10 |
tmlind | and compare to ip neighbour? | 18:10 |
bencoh | afaiu the NTF_MANAGED flag allows the userspace to avoid some entries being evicted | 18:10 |
tmlind | arp -a -n | 18:10 |
tmlind | ip neigh show | 18:12 |
tmlind | i guess it shows the same results? | 18:12 |
bencoh | pretty much yeah | 18:15 |
bencoh | anyway, maybe this interval_probe_time_ms sets a timer independently of the fact that there is no managed entry | 18:17 |
bencoh | which would explain what you saw | 18:17 |
bencoh | my view on this is that it's useless on chimaera (with no compatible userspace to use it), so ... we can just set it to some insanely high value | 18:18 |
Wizzup | bencoh: ok, do we go for -insanely high- or just really high | 18:19 |
Wizzup | like, we can set it to once every 5 or 10 minutes and it won'r really impact pm | 18:19 |
bencoh | heh, I guess just really high would work :) | 18:19 |
Wizzup | or we can set it to many years :) | 18:19 |
bencoh | (between that and the memory pressure worker that woke up every now and then and destroyed idle efforts, it really feels like they hardly care for embedded / battery-powered sometimes ...) | 18:20 |
tmlind | yeah looking at powertop neigh still runs sometimes with -1 | 18:23 |
Wizzup | tmlind: hm, with it set for everything? | 18:25 |
Wizzup | bencoh: they probably care but don't give it any thought | 18:25 |
freemangordon | Wizzup: no idea what this is about :) | 18:25 |
Wizzup | freemangordon: https://github.com/maemo-leste/bugtracker/issues/704 | 18:26 |
freemangordon | I mean - I have no clue what neigh is ;) | 18:26 |
freemangordon | will read about it, bit not now | 18:26 |
Wizzup | probably not too necessary | 18:28 |
Wizzup | tmlind: what is the process name in powertop | 18:28 |
freemangordon | is that some fancy name for ARP? | 18:29 |
bencoh | freemangordon: pretty much (it's more generic and is relevant for ipv6 as well) | 18:30 |
freemangordon | ok | 18:30 |
freemangordon | thanks | 18:30 |
freemangordon | https://elixir.bootlin.com/linux/latest/source/net/ipv4/arp.c#L166 | 18:56 |
Wizzup | hm? | 19:31 |
freemangordon | the default values, iiuc | 19:31 |
Wizzup | I still see neigh_periodic_work | 19:34 |
Wizzup | with the time set to 2147483628 for all | 19:34 |
Wizzup | sysctl -a | grep neigh | grep interval_probe | 19:34 |
tmlind | Wizzup: yeah neigh_periodic_work or something in powertop | 19:34 |
Wizzup | there is also net.ipv4.neigh.default.gc_interval | 19:34 |
Wizzup | set to 30 | 19:35 |
tmlind | no idea about that stuff | 19:35 |
freemangordon | pro1 is out of stock :( | 20:10 |
Wizzup | freemangordon: I could try to send mine if you want one | 20:51 |
freemangordon | well, I hope they will produce more | 21:15 |
freemangordon | will mail them asking what are the plans | 21:15 |
uvos | so can i issue beowulf and chimara builds at the same time yet? | 23:18 |
uvos | or is this sill an issue | 23:19 |
Wizzup | I never had issues with it personally, but I recall you or fmg did a while ago | 23:22 |
uvos | both of us | 23:22 |
uvos | we had pacakges form chimeara ending up in beowulf | 23:22 |
uvos | but i gues for the kernel this dosent matter for once xD | 23:23 |
uvos | so maybe i should try it again... | 23:23 |
Wizzup | I'm sure if we need the kernel for beowulf at this point | 23:23 |
Wizzup | I'm not sure* | 23:23 |
uvos | i mean in the kernel case its essentaly free since it cant really cause any beowulf sepcific issues | 23:24 |
uvos | besides ci issues | 23:24 |
Wizzup | sure | 23:24 |
uvos | ok i will see what happens | 23:25 |
uvos | tmlinds idle test patch is included now | 23:25 |
uvos | btw | 23:25 |
Wizzup | ty | 23:25 |
Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!