<rsalvaterra>
lipnitsk: So tomorrow I'll do a build with MIPS cpuidle and the TEO governor. :P
<rsalvaterra>
lipnitsk: So tomorrow I'll do a build with MIPS cpuidle and the TEO governor. :P
<rsalvaterra>
lipnitsk: So tomorrow I'll do a build with MIPS cpuidle and the TEO governor. :P
<mangix>
lipnitsk: btw the guy who wrote the mt7621 NAND driver is a mediatek employee. keep that in mind.
<mangix>
lipnitsk: btw the guy who wrote the mt7621 NAND driver is a mediatek employee. keep that in mind.
<mangix>
lipnitsk: btw the guy who wrote the mt7621 NAND driver is a mediatek employee. keep that in mind.
<lipnitsk>
mangix: well right... so is the guy who wrote the one in the kernel. Then Greg came in and wrote his version for whatever reason, which has references like:
<lipnitsk>
mangix: well right... so is the guy who wrote the one in the kernel. Then Greg came in and wrote his version for whatever reason, which has references like:
<lipnitsk>
mangix: well right... so is the guy who wrote the one in the kernel. Then Greg came in and wrote his version for whatever reason, which has references like:
<lipnitsk>
mangix: I see this from Sergio - "MIPS: ralink: add CPU clock detection and clock driver for MT7621"
<lipnitsk>
clock != memory
<lipnitsk>
mangix: while we are on the subject, you wouldn't happen to have a rt288x device lying around or to explain why we need "ramips: Fix early memory calculation for certain MIPS platforms" patch?
<Namidairo>
Grommish: as in that device does not have lzma-loader
caiortp has quit [Remote host closed the connection]
<Grommish>
Ah
<Grommish>
Made me panic about rebasing for a minute :)
<Namidairo>
I just don't remember what size the kernel had to get to before the stock bootloaders started breaking
damex has quit [Read error: Connection reset by peer]
damex has joined #openwrt-devel
<Grommish>
Namidairo: Ah, my image is 4.23 MB (4,435,968 bytes) in size, but uses the lzma compression
<Rene__>
Grommish: on your UBNT ER device?
<Grommish>
Rene__: Yes
<Grommish>
Rene__: 4435059 Mar 16 05:26 openwrt-ramips-mt7621-ubnt_edgerouter-10x-initramfs-kernel.bin
<Rene__>
I haven't seen any issue on my UBNT ER device. But I also have a Strong 1200, with is a mt7621 wifi router. Which does has issues
<Grommish>
Rene__: *nod*
<Grommish>
MT76 has been a adventure and experience for me for sure
<Rene__>
booting via tftp is not the issue. It happens when booting from flash
<Grommish>
I'm really glad I don't have to deal with the wifi aspects
<Rene__>
:)
<Grommish>
Rene__: I need to figure out the LEDs on the er10x and try and fix the single port that refuses to work.. Then maybe PoE if I feel up to it hehe
koniu has quit [Remote host closed the connection]
koniu has joined #openwrt-devel
<Rene__>
normally passive PoE should not influence the port workings. PoE signal is injected via the ethernet transformer.
<Rene__>
Which ethernet port of which switch?
<Grommish>
switch0:port0 on the rtl8367rb
<Grommish>
It's like its disabled, but the other 4 ports (switch0:1-4) work fine
<Grommish>
I post something in the Networking area of the forum. I suspect it's a swconfig issue, but I've never used swconfig
<Rene__>
ok
<Grommish>
and mixing swconfig and dsa makes people go Ewwww
<Grommish>
but can't be helped right no
<rsalvaterra>
Grommish: So you need swconfig for the rtl8367rb and DSA for the other switch?
<Grommish>
rsalvaterra: Yep
<rsalvaterra>
Ugh… what a mess.
<Grommish>
Indeed, but it works
<Grommish>
except for that one port hehe
<Grommish>
This thing takes the entire rtl switch as a whole and funnels it into the mt76 board
<Grommish>
Downside is that you can't address each as its own eth port I guess, which is a pain
<Grommish>
So the switch is labeled as "eth5" in ifconfig
<mangix>
Given that the totolink purchase fell through, I bought a device that I will need to add support for
<mangix>
mt7621 i think
<Grommish>
mangix: Ooo.. Whatcha get?
<Grommish>
Fresh-electronic device smell.. Yum
<mangix>
Linksys E7350
<Grommish>
mangix: Nice!
<mangix>
looks like similar enough hardware
<mangix>
I hope adding support for it is boring
<Grommish>
mangix: True.. Boring is underrated hehe
<mangix>
I've only written support for one device, which never got sold.
skolev has joined #openwrt-devel
<mangix>
rsalvaterra: turns out armbian uses zram by default on half the RAM. slowed my NAS to a crawl
victhor has joined #openwrt-devel
<rsalvaterra>
mangix: It was badly configured, then.
<rsalvaterra>
What compression algorithm is it using? And what's the maximum amount of memory it's allowed to allocate?
<mangix>
half of the RAM. In my case, 2GB -> 1GB
zatwai has quit [Quit: ZNC 1.8.2+deb2~bpo10+1 - https://znc.in]
<mangix>
slows it down when using above 1GB
zatwai has joined #openwrt-devel
gch9812132 has joined #openwrt-devel
<rsalvaterra>
Half might be too much, depending on the compression algorithm.
<mangix>
I would guess it's LZO. No idea.
gch981213 has quit [Read error: Connection reset by peer]
gch9812132 is now known as gch981213
<rsalvaterra>
cat /sys/block/zram0/comp_algorithm
<mangix>
gch981213: welcome. lipnitsk was wondering about upstreaming about your mt7621 memory patch
<mangix>
cat: /sys/block/zram0/comp_algorithm: No such file or directory
<rsalvaterra>
mangix: Of course, if you're not using zram… :P
<mangix>
it's in use currently
<rsalvaterra>
Impossible.
<rsalvaterra>
If you're using zram, you have at least one device (zram0).
<rsalvaterra>
We have per-CPU compression streams. None of this nonsense of creating as many zram devices as CPUs.
<rsalvaterra>
mangix: This is what I have in my Debian/Ubuntu systems. A heavily modified version of the zram-config package. https://paste.debian.net/1189702/
<svanheule>
Raheel: try #openwrt or the forums with this type of question, and provide some details (which device, what is "too slow")
<Raheel>
svanheule thanks for the reply. Too slow means i am doing zoom call and streaming with wired client but its not working fine but as i remove the bridge interface of lan eth1.1000 then it works fine and no breaking in zoom calls
gch981213 has quit [Read error: Connection reset by peer]
gch981213 has joined #openwrt-devel
Raheel has quit [Quit: Connection closed]
<gch981213>
mangix: lipnitsk: I was planning on upstreaming that memory patch after the clock one. I wasn't able to get the clock patch upstreamed back then so this memory detection just stayed in openwrt till now.
<gch981213>
lipnitsk: about mt7621 nand driver: Weijie said elsewhere that the nand controller is similar to the mt7622 one, but the one in mt7621 got broken DMA burst support.
<lipnitsk>
gch981213: thanks for the info. hope you don't mind me trying to upstream the mem detect patch then.
<gch981213>
lipnitsk: datasheet for mt7622 is available from Banana Pi R64 wiki, and mt7621 datasheet can be obtained by googling 'mt7621 programming guide' There's a closed PR on GitHub adding DMA support for current mt7621 nand driver. (I rejected that one due to an unnecessary oob layout change.)
<lipnitsk>
gch981213: so there is a way to use DMA without burst mode on mt7621?
<lipnitsk>
rsalvaterra: great news on the cpuidle. any downsides to just enabling it?
<gch981213>
lipnitsk: Mediatek has at least 4 different bad block table implentations. And those will definitely be rejected upstream.
<rsalvaterra>
lipnitsk: If you're reading this, it's working fine. :)
<lipnitsk>
rsalvaterra: does the device run cooler with it or something? less power?
<jow>
gch981213: rsalvaterra: thank you
valku has joined #openwrt-devel
<gch981213>
lipnitsk: They somehow don't want to use UBI till recently so they need something else to deal with bad blocks.
<rsalvaterra>
lipnitsk: It should use less power, yes, but I haven't measured.
<jow>
gch981213: rsalvaterra: could you run this too? "grep DEVTYPE /sys/class/net/*/uevent" - background is that I look for a good way to discover DSA ports, blogic just gave me the tip to check uevent, which shows DEVTYPE=dsa
<gch981213>
lipnitsk: I'd prefer letting the patch staying in master for a while before backporting. I'll do that next week.
<jow>
rsalvaterra: the ports denoted with '*' will set the "option vlan" value as pvid
<lipnitsk>
gch981213: okay, thank you, that sounds good to me. just making sure it wasn't forgotten.
<jow>
so lan1, lan2 will get pvid 100 and lan4 will get pvid 200 in the pasted example
<jow>
if a port is marked with '*' in multiple vlans, the behaviour is undefined, afair the last one takes precedence then though, due to the way the config is parsed
<jow>
and it could be that line 4 of my paste is redundant since the union of all brdige-vlan ports is added as bridge port members
<rsalvaterra>
jow: Ooooh, I had seen that somewhere in your tree, but I though it hadn't been merged yet!
<jow>
it has been added into netifd
<jow>
the stuff in my tree was a shell based mockup
<jow>
you can indeed omit the entire config device section if you target the preexisting br-lam
<jow>
*br-lan
<jow>
so merely adding config bridge-vlan sections with "option device br-lan" should suffice to enable vlan filtering
<rsalvaterra>
Great, that means we're on par with swconfig, LuCI being the only missing piece. :)
<jow>
yeah, still trying to fix various ugly corner cases
<rsalvaterra>
We could just… you know… deprecate LuCI…
* rsalvaterra
runs
<jow>
yeah great plan. break backwards config, introduce dsa without config port and then throw away the entire ui because stuff is hard
<jow>
users going to love it
<rsalvaterra>
I'm obviously kidding… ;)
<jow>
I guess we should deprecate uci too and just ship a huge rc.local with iproute2 commands, people can just edit it to their needs
<jow>
I know :)
<rsalvaterra>
jow: I did a test yesterday… built an image for my WDR3600 selecting only zstd as the jffs2 compression algorithm (disabling both lzo and rtime). It cut about 200 kB from my sysupgrade image.
<rsalvaterra>
*lzma and rtime
<olmari>
sounds promising
<jow>
nice
<rsalvaterra>
Really… lzma is a pig.
guidosarducci has quit [Remote host closed the connection]
<rmilecki>
jow: "sw" is external switch handling 4 extra LAN ports
<Grommish>
Interesting.. I was having issues with suricata not finding pcap.h. I was setting a sysroot and pointing it directly at the include and lib dir.. Turns out, the <sysroot>/usr/lib/pcap.h just pointed to the host pcap.h.. which isn't listed as a build enviroment pre-req.. usr/lib/pcap/pcap.h is what needs to be used
<Grommish>
I wonder why the usr/lib/pcap.h doesn't point to the usr/lib/pcap/pcap.h rather than <include/pcap.h>
<guidosarducci>
jow: hey there, wondering if you're OK pulling in the fw3 DSCP fixes? Let me know if you're getting swamped and I'll squash and try sending to the mailing list. Cheers!