abstracttim | Hi people | 00:31 |
---|---|---|
clort | usb networking is rather nice but if i put two of the same devices on the hub, i get usb 1-2.3.3 "Device not responding to setup address" | 01:29 |
clort | more bluetooth ickyness DSA-4774-1 https://lists.debian.org/debian-security-announce/2020/msg00181.html | 07:51 |
tuxd3v | clort, yeah linux is starting to become a hot spot unfortunately | 07:59 |
clort | le wut. this stuff happens. | 08:00 |
tuxd3v | not only that, at each time linux kernel grows, and grows, even selecting the same config, you get always a bigger kernel image.. | 08:00 |
clort | i'm a floofy cat with blacklisted bluetooth modules since it was introduced | 08:04 |
clort | bluetooth oopsies are pretty regular occurrences | 08:05 |
tuxd3v | megi: I tested your patch to fix ethernet on reboot, it seems to be working :) | 08:40 |
tuxd3v | megi: but only if "rgmii-id" is used in the DT.. | 08:42 |
clort | could we have iptables back, for devuan ceres? | 09:42 |
clort | <insert random angry language here> | 09:42 |
clort | i need to iptables -A POSTROUTING -t nat -s 192.168.42.2/32 -j MASQUERADE | 09:43 |
Joril | clort: It looks like the iptables package is still there...? | 09:50 |
Joril | it contains iptables-nft, that should allow you to use the iptables syntax | 09:51 |
clort | ptables -A POSTROUTING -t nat -s 192.168.42.2/32 -j MASQUERADE | 09:55 |
clort | iptables: symbol lookup error: iptables: undefined symbol: xtables_fini | 09:55 |
clort | iptables-nft -A POSTROUTING -t nat -s 192.168.42.2/32 -j MASQUERADE | 09:56 |
clort | iptables-nft: symbol lookup error: iptables-nft: undefined symbol: xtables_fini | 09:56 |
clort | iptables-nft -L | 09:56 |
clort | iptables-nft: symbol lookup error: iptables-nft: undefined symbol: xtables_fini | 09:56 |
clort | it does help me to know that it 'should work' though Joril - thank you | 09:57 |
clort | do you think it might be related to using nvidia's 4.9.140-tegra kernel? | 09:58 |
Joril | ouch! Found this https://www.spinics.net/lists/netfilter-devel/msg68446.html | 09:58 |
Joril | But it looks like no one answered | 09:58 |
Joril | It could be that some other package related to iptables isn't sufficiently up-to-date... libxtables maybe? | 10:04 |
gnarface | if it's a kernel you got from nvidia, really don't necessarily expect any competence, but don't rule out utter malice | 10:05 |
gnarface | but it could be just a module that didn't auto-load too | 10:05 |
gnarface | so maybe check for anything missing? | 10:06 |
gnarface | if it's packaged right, the modules would all be in /lib/modules/ somewhere | 10:06 |
gnarface | (/lib/modules/`uname -r`, specifically, but you can't trust nvidia to follow the rules that well) | 10:07 |
clort | i see they have nf_tables.ko and nf_tables_set.ko and not iptables. maybe i can build the 4.9.140 module myself and dynamically load it | 10:18 |
clort | or maybe some smarty person can make a 5.8 kernel dtb for my hardware (jetson nano) that i can boot | 10:19 |
gnarface | the dtb you have already won't work? | 10:23 |
gnarface | or they embedded it in the binary so you can't get it? | 10:24 |
gnarface | building the modules and loading them might work | 10:25 |
gnarface | it might also work to kexec 5.8 from 4.9 | 10:25 |
clort | even funnier | 10:26 |
clort | modprobe: FATAL: Module nf_tables.ko not found in directory /lib/modules/4.9.140-tegra | 10:26 |
clort | ✠ /lib/modules/4.9.140-tegra# ls -l nf_tables.ko | 10:26 |
clort | -rw-r--r-- 1 root root 216472 May 25 11:35 nf_tables.ko | 10:26 |
clort | wtg nvidia | 10:27 |
clort | they don't supply a newer dtb / kernel for me | 10:27 |
clort | if you'll allow me to remark, not allowing the kernel module you ship to load isn't brilliant | 10:46 |
clort | * nvidia | 10:46 |
clort | i hoped i'd be okay with dropping their ubuntu in favor of devuan, but clearly this is not enough | 10:46 |
clort | their gpu and opengl support is best available, for a $99 ARM SBC though, so for that I'm happy | 10:48 |
gnarface | the kexec trick might work, worth looking into | 10:56 |
gnarface | make sure it won't brick it though | 10:57 |
unixbsd | Hello, I am using samba on raspberry pi zero. I have impression that samba is faster than NFS and sshfs. is it possible? | 12:55 |
gnarface | faster than sshfs would be expected, faster than nfs sounds like something is wrong | 13:10 |
gnarface | but it could be a difference in caching settings | 13:11 |
gnarface | or a number of other bottlenecks i could imagine | 13:11 |
gnarface | you've ruled out issues related to flash storage, slow bus speeds, and not enough ram? | 13:12 |
gnarface | the thing has so little ram that i could imagine a scenario where just accessing the files by samba first would cripple nfs if you repeated the test immediately there afterwards from the same device | 13:13 |
gnarface | maybe try from a clean boot, and try it from nfs without samba running | 13:14 |
clort | you should be able to pick a low-overhead cipher for sshfs also | 13:18 |
gnarface | yea, but nfs shouldn't be dragging i don't think... | 13:39 |
gnarface | unless there was some problem in the broadcom firmware | 13:39 |
gnarface | but yea using unencrypted sshfs would be a slick way to avoid dealing with it | 13:40 |
rrq | mmm "icherboot" | 14:31 |
tuxd3v | clort, is the support for jetison nano availlable in mainline? | 16:05 |
tuxd3v | I mean also uboot? | 16:05 |
tuxd3v | let me check | 16:05 |
clort | i dunno tuxd3v | 16:07 |
tuxd3v | nope, in linus mainline is not present :( | 16:09 |
tuxd3v | you need to get a kernel from: | 16:09 |
tuxd3v | https://developer.nvidia.com/embedded/downloads | 16:09 |
tuxd3v | clort, its possible, but only with nvidia sources | 16:19 |
tuxd3v | A devuan rootfs its also possible and easy :) | 16:21 |
clort | thank you tux3dv - that would get me drivers for say, a cdrom, or iptables ? | 16:22 |
tuxd3v | clork, with the kernel sources then you have to 'make menuconfig', and choose the drivers you want | 16:23 |
tuxd3v | iptables, cdrom and so on | 16:24 |
tuxd3v | but its a tedious task | 16:24 |
clort | it's unfortunate nvidia management decided to support the Nano as an ai-appliance only, rather than a general purpose little computer. | 16:32 |
clort | however with mainline kernel and usb OTG, along with a powered hub, a Droid4 might serve as my internet router with a usb-ethernet adapter | 16:36 |
clort | along with cdrom support ;) | 16:36 |
hightower2 | Hey, does anyone know (just offhand, not as a long discussion) why devuan offers the choice of linking bin to usr/bin, rather than the other way around? (that is, /usr/bin being a link to large /bin)? | 18:26 |
DHE | historically /bin was meant to be on / which could be a small(ish) filesystem and contain mainly system-specific and recovery applications. /usr could be a shared NFS mount for many servers. install once, available everywhere | 18:30 |
DHE | *system-specific configs | 18:30 |
danuan | trying to run a backup script from cron, but having trouble sourcing subscripts from it without full path to them. does bash not look for (source somescript.sh) in $PATH ? | 18:31 |
hightower2 | DHE right, sure, historically. But I mean, when these are joined, why the decision was made that "/usr/bin" was the real one and "/bin" just a symlink to it, rather than the other way around? It would seem to me that now, when / and /usr separation is no longer relevant, things get moved from /usr/ into /, and not from / into /usr/. | 18:34 |
clort | does cron invoke executeables with bash danuan ? | 18:34 |
hagbard_ | If the script is in $PATH you can execute it. But not source it. | 18:34 |
danuan | clort if executable has #!/bin/bash should it not ? | 18:35 |
danuan | hagbard_ so i have to cd in to the running dir within the script before sourcing or set a local $mypath | 18:36 |
hightower2 | danuan, if you already know which directory it's in, then don't 'cd' into it, but simply execute source /path/to/dir/script.sh | 18:36 |
clort | i was imagining cron might exec a script and pass it literally to bash to interpret, without allowing bash to scan the whole path. but i don't know. | 18:37 |
hagbard_ | yes. Or hashbang the script, make it executable, and execute it. | 18:37 |
hightower2 | (if you really needed to run 'source'... running 'source' from cron seems pretty unusual) | 18:37 |
hightower2 | ah you're running subscripts.. ok | 18:37 |
danuan | i am not sourcing from cron itself , i am sourcing from within a bash script cron runs | 18:38 |
hightower2 | (even though - same question... running "scripts" using 'source' is also pretty unusual) | 18:38 |
danuan | hagbard_ it is part of the script , cannot execute , need to pass variables in and out . without doing extra $1 $2 $3 s | 18:40 |
hagbard_ | Then you might need indeed a cd somewhere. | 18:41 |
danuan | ok thanx, thought that since the subscripts were executable and .sh it would look in $PATH but guess not for sourcing | 18:45 |
ErRandir | The behaviour of source depends if bash is in posix mode. See man bash. You may also want to put 'SHELL=/bin/bash' in the crontab file | 18:47 |
ErRandir | The default shell used by cron is /bin/sh | 18:48 |
danuan | ErRandir but even if it is /bin/sh but you ask it to execute #!/bin/bash script it will run is /bin/sh ? | 18:50 |
hightower2 | in that case it'll run bash | 18:52 |
ErRandir | I do not think cron will look at that. It will apply to any subshell created from the script. | 18:53 |
hightower2 | but folks, wait... if this guy has a script, which has #!/bin/bash at the top... then if he calls 'source' from this script, the sourced file won't be processed by anything other than that same (bash) process itself. | 18:54 |
ErRandir | it's not a bash process, it's a sh process. | 18:54 |
hightower2 | if he runs /path/to/script.sh from cron, and script has #!/bin/bash at the top, it'll execute as bash | 18:56 |
hightower2 | (the main script I mean) | 18:56 |
danuan | ErRandir i wrote my script as a bash one , it would error out a long time ago if it was executing in /bin/sh | 18:57 |
danuan | and changing crons shell , i would assume woult throw the all the systems cron task in to a dissary i would assume ? would it not ? | 18:58 |
ErRandir | so then you must indeed be in bash for the script | 18:58 |
lyubov | ~/buffer 33 | 19:12 |
fsmithred | clort, what's a low overhead cypher now that blowfish is gone? | 19:36 |
clort | ick. maybe arcfour according to websearch | 19:38 |
clort | i shall test | 19:38 |
fsmithred | I don't see that in the list generated by 'ssh -Q cipher' | 19:39 |
hagbard_ | I'd assume that aes is quite low overhead, since modern cpus have hardware support for it. | 19:39 |
fsmithred | oh, so we have to test on every machine | 19:40 |
clort | aes128? | 19:41 |
fsmithred | maybe. The 128 is a good sign. | 19:41 |
fsmithred | they are not in alphabetical order | 19:42 |
fsmithred | so maybe they get stronger as you go down the list? | 19:42 |
clort | well to my phone i get 4.3MB/s with aes128-ctr and 4.1MB/s with aes128-gcm@openssh.com | 19:47 |
fsmithred | I got 95MB/s and 109MB/s with those same two, respectively. They reversed. | 19:54 |
clort | going the other way gcm@openssh.com is 6.4% faster | 19:55 |
fsmithred | for aes256-gcm@openssh.com I got 107-108MB/s. Almost the same as the 128. | 19:57 |
fsmithred | FWIW these are both core i5 about 10-11 years old. | 19:57 |
clort | i should have monitored cpu more closely, neither time was i saturating one core | 19:58 |
fsmithred | 110MB/s if I don't specify which cipher to use. | 19:59 |
fsmithred | so, no significant difference. I don't need to do the extra typing | 20:00 |
fsmithred | thanks for testing | 20:00 |
clort | i try to cat file | nc to another computer but it doesn't end at eof | 20:11 |
clort | ah -q 0 | 20:13 |
clort | reading from microsd, sending to ssd, over usb3 network with netcat i get 11.1MB/s | 20:15 |
clort | reading from microsd to dev null i get 20.9MB/s | 20:20 |
deadrom | hi | 20:24 |
deadrom | how different is devuan from debian, how alike? is stable production stable? | 20:27 |
mason | deadrom: Very similar. Most packages are built by Debian, except those that have to be forked because they require systemd. | 20:28 |
ErRandir | my system is definitely production stable. My only time I do a reboot is when there is a power outage. | 20:30 |
clort | too many variables between use-cases for blanket statements | 21:26 |
clort | i haven't noticed any drawbacks vs debian for myself | 21:27 |
sgage | deadrom: one caveat re: repos. Debian's 'testing' becomes 'stable' | 21:31 |
sgage | some time before Devuan. E.g., currently, Devuan testing is tracking | 21:32 |
deadrom | umm, no, did not get that | 21:32 |
sgage | Debian testing (bullseye) (Devuan testing is chimaera) | 21:32 |
sgage | When bullseye becomes stable, chimaera will still be tracking it, | 21:33 |
sgage | although it will still be Devuan's testing. | 21:33 |
sgage | Long and short - use the names (ascii, beowulf, chimaera) in | 21:34 |
deadrom | but Devuan has its own repos it sticks too? | 21:34 |
sgage | sources.list instead of stable/testing | 21:34 |
deadrom | ok | 21:34 |
sgage | Devuan has its own repos, but if apt requests a file that did not in | 21:34 |
sgage | any way have to be tweaked to work with systemd, the magical amprolla | 21:35 |
sgage | gets it from the Debian repo. If the package has -devuan in the name | 21:36 |
sgage | it has been forked by Devuan and is on their own servers. | 21:36 |
sgage | plenty of info on devuan.org and the forum. If you | 21:37 |
sgage | want to experiment with Devuan, it is worth understanding the model | 21:37 |
sgage | of how the packaging system works. | 21:37 |
deadrom | ok, thanks | 21:39 |
golinux | deadrom: How devuan repos work: https://dev1galaxy.org/viewtopic.php?id=3192 | 21:41 |
danuan | wondering if it would be safe to dd backup a mounted system disk from singleuser runlevel. usually i boot off usb to do a full system dd image , but what about automatic switch to runlevel 1 remount ro run dd and back to normal runlevel | 21:58 |
danuan | and yes i do rsync also , but its nice to have a ready to go identical disk | 21:59 |
fsmithred | danuan, you have to exclude some things if you want to boot. | 22:01 |
fsmithred | anything that doesn't exist when the machine is turned off should be excluded | 22:01 |
fsmithred | if you want to dd the whole drive, boot from usb like you already do | 22:02 |
danuan | exclude in a dd ? and point is not to turn off the machine , do it while it is running | 22:02 |
fsmithred | no, exclude with rsync | 22:03 |
danuan | ohh yes i know | 22:03 |
fsmithred | or exclude with tar | 22:03 |
fsmithred | or dd a hibernated image? | 22:03 |
fsmithred | but that involves off | 22:03 |
danuan | but the actuall question, does that seem safe ? dd a readonly mounted system ? ive never done dd on mounted running system | 22:05 |
fsmithred | How will you bring it up if you ever need to use that image? | 22:06 |
fsmithred | It already thinks it is up and running. | 22:06 |
danuan | to backup, without shutting down | 22:07 |
danuan | not to replace a running system | 22:07 |
fsmithred | you just want to back up data? | 22:07 |
fsmithred | not the OS? | 22:07 |
danuan | no the whole system disk | 22:08 |
danuan | backup with dd while it is running | 22:08 |
fsmithred | and do you expect to ever use that backup image to restore to disk and then run that system? | 22:08 |
danuan | yes | 22:08 |
fsmithred | how can you boot an already running system? | 22:08 |
fsmithred | that's what your image will be | 22:09 |
fsmithred | I'm talking about after you restore the image | 22:09 |
danuan | backup running system without shutdown , and off course reboot when need to restore | 22:09 |
fsmithred | as it starts to reboot, it will see stuff that should not exist | 22:10 |
fsmithred | I think you'll need to back up the contents of memory and restore that, too. I don't even think that's possible. | 22:10 |
danuan | ahh i think i get it , i am trying to backup a running state | 22:10 |
fsmithred | yeah | 22:10 |
fsmithred | all of /sys /proc /dev and /run should be empty | 22:11 |
fsmithred | and I'm not sure what else | 22:11 |
fsmithred | a different approach would be to make a bootable live-iso of the system using refractasnapshot | 22:12 |
fsmithred | and include refractainstaller, so you can install it if you need to restore | 22:12 |
fsmithred | back up all your data separately, so you don't make a huge iso | 22:12 |
fsmithred | that way, it doesn't even need to be restored to the same hardware. | 22:13 |
danuan | ok, i get it , not the greatest idea | 22:13 |
danuan | will stick with periodic dds from shutdown state and rsyncs to get back to latest | 22:13 |
fsmithred | I've done the dd method for a whole drive, and it is nice. Just attach and boot the copy if you need to. | 22:13 |
danuan | yea its great untill you leave both drives attached and then it can mount some partitions from one drive soem from other as uuids match :) | 22:14 |
fsmithred | what you're doing makes sense. OS doesn't change that often. | 22:14 |
fsmithred | lol | 22:15 |
fsmithred | I found out that happens with lvm, too | 22:15 |
fsmithred | if the names match | 22:15 |
fsmithred | I need to go outside. bbl. | 22:16 |
n4dir | whenever i backup with the rsync command from arch.wiki, it rsyncs "swap" file too (swap iirc). I don't remember that exactly though. Else their excludes seem to be what does it | 22:22 |
n4dir | whenever i backup with the rsync command from arch.wiki, it rsyncs "swap" file too (swap iirc). I don't remember that exactly though. Else their excludes seem to be what does it | 22:22 |
n4dir | whenever i backup with the rsync command from arch.wiki, it rsyncs "swap" file too (swap iirc). I don't remember that exactly though. Else their excludes seem to be what does it | 22:22 |
n4dir | oh my god. Sorry folks | 22:22 |
syco- | man rsync | 22:29 |
danuan | exclude paths are relative , i ran it to that rsyncing home dirs trying to exclude .cache , needs a bit of thinking | 22:30 |
danuan | relative to current things you are trying to rsync | 22:30 |
danuan | fsmithred ps dd is not rsync so i think that /sys /proc /dev/ /run would not be backed up as they are not on the actual hard filesystem just virtual one | 22:32 |
n4dir | ah yeah, cache, that sucks too | 22:40 |
n4dir | syco-: why would man rsync help with that, or with what you think it might help? | 22:41 |
fsmithred | danuan, interesting idea. I've never tried it. | 22:50 |
danuan | sound like a perfect test for qemu-kvm a script that switches runlevel , remounts ro does a dd backup to a drive or an image and change runlevel back again , and then see if that image would boot normaly | 22:52 |
frej00 | it should be utterly fine | 22:52 |
frej00 | It would be the same thing as a sudden power outage, and then booting up the machine later | 22:53 |
danuan | yesnot a sudden as all caches would flush normaly when going ro | 22:54 |
fsmithred | does it even need to be remounted ro? | 22:55 |
frej00 | it would be wise | 22:55 |
frej00 | otherwise things could be inconsistent | 22:56 |
danuan | yes , otherwise you could just dd a fully running system and hope for fsck to fix it later from midwrite files and such things | 22:58 |
danuan | fsmithred and frej00 , just tried it under qemu-kvm , seems to have worked just fine , boots to a new image without problems | 23:34 |
fsmithred | cool | 23:34 |
frej00 | sweet | 23:34 |
danuan | makes me wonder now , for shits and giggles while it is under ro is it possible to switch /dev/drive1 to /dev/drive2 | 23:35 |
fsmithred | you mean like hotplugging drives? | 23:36 |
danuan | like after you do dd copy while in single user , switch the running system from one drive to the other and bring it back up to normal runlevel | 23:37 |
danuan | just for experiments sake | 23:37 |
fsmithred | how do you switch which drive you're running? | 23:37 |
danuan | do not know yet , but interisting to see if i can | 23:38 |
DHE | the command is pivot_root however programs running from the old drive are still running from the old drive. might be a problem with pid 1 and require an init reload | 23:38 |
danuan | but running while mounted ro in single user and after dd copy drives are identical | 23:39 |
danuan | so it is possible to switch system drive on a running system then ? | 23:39 |
DHE | yes, but it's a fair amount of work on its own | 23:39 |
onefang | There will be a bunch of things with open files on the old system drive. | 23:58 |
DHE | hopefully in single user mode there's little more than init (which should be reloaded) and the shell you are using for single user mode which can also be reloaded. but anything else will likely be a concern | 23:59 |
danuan | cannot be that easy i just typed mount /dev/sb1 / and it overmounted old /dev/sda1 | 23:59 |
Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!