IRC log of #cubox of Sat 15 Feb 2014. All times are in CET < Back to index

07:43 scattym Hey, anyone around that can help me migrate back to lircd from eventlircd on the geexbox build?
07:47 jnettlet sorry I have not played around with either on the new builds.
07:54 scattym no probs.
07:55 scattym the guides I have followed so far don't seem to show which lircd driver I should use to point at the internal ir port, only for adding an external IR receiver...
08:30 warped-rudi_ scattym, the internal IR port is enabled by default on geexbox
08:32 warped-rudi_ jnettlet, speed testing with only one CPU gave no difference: HB->150MB/s i4pro->100MB/s
08:33 jnettlet is it possible that the cpu-scaling is working differently on one vs the other.
08:33 jnettlet there are a lot of factors that could play into this.
08:37 warped-rudi_ hmm, any ideas where to look ? I'm using Rabeeh's 3.0.35 kernel with some tweaks...
08:38 jnettlet cat /proc/cpuin
08:38 jnettlet cpuinfo
08:39 jnettlet but you can use the sysfs interface to limit both to the same frequency for testing
08:41 scattym warped-rudi, thanks. I am having trouble with implementing the apple remote and was hoping to move to lircd where I had a little more control over the IR params
08:41 scattym had = have
08:44 warped-rudi interesting, the BogoMIPS change between 790 and 1988 on the i4. let's look what it does on the HB
08:46 scattym oh, looks like you're in the middle of something. I'll check back later.
08:50 warped-rudi scattym, I think switching to lirc is kinda documented on the geexbox web site. However, I don't know if this will work for the internal IR.
09:02 warped-rudi BogoMIPS on the HB bounce between the same values. but less frequently.
09:07 mk01 1990 is ok per cpu (on 1ghz)
09:31 scattym yeah, that's where I got stuck. I had been over the documentation on the web site. seems more targeted at an external ir receiver. will keep plugging away. thx.
09:37 jnettlet warped-rudi, you are using the latest SPL based u-boot right?
09:38 jnettlet I am getting 276 MB/s cached reads on my i4p
09:40 warped-rudi Yes, latest commit from https://github.com/rabeeh/u-boot-imx6/
09:41 warped-rudi BTW, the reason why I noticed that something might be wrong is that the flow of messages from systemd is visibly slower on the i4 than on the HB.
09:43 jnettlet warped-rudi, try this u-boot and see if there is any difference. https://dl.dropboxusercontent.com/u/736509/u-boot/uboot_debug.tgz
09:44 jnettle 09:44 * jnettlet is going to walk the dogs now
09:51 mk01 warped-rudi: internal IR should work with LIRCD with the help of "ir_lirc_codec" keymap.
15:27 matoking Hey, anyone have an idea why my upload speed starts at 600 KB/s and then goes down to around 100 KB/s?
15:27 matoking I've tried SFTP, FTP using vsftpd and Samba and all work like that
15:28 matoking I'm using an USB HDD and I got a write speed of about 20 MB/s
15:28 mk01 upload to Cubox ?
15:28 matoking Yes
15:28 mk01 20MB/s is tested how ?
15:29 matoking dd if=/dev/zero of=testfile0 bs=8k count=10000;
15:29 matoking It's ridiculous since I can download from the Internet at around 800 KB/s at best
15:30 matoking It's either my Linux laptop is doing a shoddy job somehow or my router is a piece of junk
15:30 mk01 it is Eth connected or Wlan ?
15:30 matoking WLAN
15:31 mk01 iwconfig wlan0 is showing what ?
15:32 matoking @mk01 http://pastebin.com/fzucte1U
15:32 matoking To be fair I've had hard times with the router
15:33 matoking For example, refreshing the server list in TF2 would bring down the Internet connection for a few minutes
15:33 matoking And now that I'm running a P2P application on the Cubox-i the connection has a higher latency even if it's all happening on the local network
15:36 matoking Hmm
15:36 mk01 with 802.11g (max 54MBps) you get ~2MB/s in real life.
15:36 matoking Yeah, that's about the upload speed I get when I upload from my Linux laptop to my phone over FTP
15:37 matoking Although it too drops a bit
15:37 matoking But yeah, I did a bit of googling and most people seem to agree
15:37 matoking This router *is* a piece of junk
15:37 mk01 it all depends on other networks, noise etc
15:37 mk01 change the router for N capable.
15:38 mk01 you can investigate more by turning off powersave mode (you have on) and "iw scan
15:38 mk01 "iw scan". you will see networks around you
15:38 mk01 overlapping channels / etc
15:38 matoking Oh
15:39 matoking Power management
15:39 matoking How can I turn it off?
15:39 mk01 iwconfig wlan0 power off
15:39 matoking Well that's a bit ambigious isn't it :P
15:40 matoking Started at 1.2 MB/s and dropped down to around 80 KB/s :/
15:41 matoking I'll have to do some testing
15:41 mk01 then check neighbour networks
15:41 mk01 iwlist scanning
15:41 matoking There is only one network around here
15:41 matoking It doesn't pick up anything else
15:41 mk01 or "iw wlan0 scan"
15:42 mk01 try to set open network
15:42 mk01 so no WEP/WPA/WPA2
15:42 mk01 (just for test)
15:42 matoking It's already open
15:42 mk01 ah
15:42 matoking (I know that's a bad idea)
15:42 matoking But I know to use SSL if I'm dealing with anything sensitiv
15:42 matoking So no worries about that
15:43 mk01 what channel is your net and the one you can see around ?
15:43 matoking Channel 6
15:44 matoking There are no other Wi-Fi networks around here besides the one I'm using right now
15:45 mk01 then it is straightforward. change router for Ncapable or connect Cubox with lan
15:45 matoking Yeah
15:46 matoking A FYI for other people who may be listening
15:46 matoking Don't buy ZyXel routers
15:47 mk01 Buy one with chipset running linux
15:47 matoking For dd-wrt?
15:48 mk01 dd-wrt / openwrt ... yes
15:48 matoking Hmm
15:48 mk01 sometimes its the FW what is crap. even with solid HW and price.
15:50 mk01 (if you have $$ to waste buy AirPort Capsule or Extreme station ;) ). you will get 20-25MB/s.
15:52 matoking I'd rather use something I can hack as I wish :)
15:53 _rmk_ matoking: it could be that the sdhci driver is causing the slow down, or the linux mmc layer. I've been going through it in 3.14-rc2 during the last week fixing lots of stuff there.
15:53 _rmk_ matoking: if the linux mmc layer hasn't been told that the host driver can signal sdio interrupts, then it'll poll for them instead
15:54 matoking @_rmk_ I'm using an USB HDD, does that matter?
15:54 _rmk_ matoking: also, if the sdhci driver does runtime PM, that can have an adverse impact too
15:54 _rmk_ the wireless is connected to a sdhci interface
15:54 matoking Oh
15:54 matoking Hmm
15:55 matoking I'm using the debian-jessie build
15:55 _rmk_ I can't speak for the freescale 3.0.35 kernel since I work on mainline stuff
15:55 matoking I suppose this fix will be as easy as installing a new kernel?
15:55 matoking Provided it's fixed, that is :)
15:55 _rmk_ and I've no idea what debian are doing with their kernel for imx6
15:55 _rmk_ I haven't pushed my changes anywhere yet because it's work in progress
15:56 _rmk_ what I'm saying is I've already found quite a bit of yuck in the v3.14-rc2 version
15:56 matoking 3.0.35-g0f2006c-dirty
15:57 matoking Yeah, my kernel doesn't seem to be exactly up-to-date :P
15:57 _rmk_ sounds like it may be derived from the freescale kernel
15:57 matoking Ah
15:57 _rmk_ freescale's official BSP is based off 3.0.35
15:58 _rmk_ to say whether any of this stuff is relevent, I'd need to dig into that tree
15:59 _rmk_ I'm at 27 patches for it, two from someone else, and the more I look at the driver, the more patches I have.
15:59 matoking Do you have a repository with those patches?
16:00 _rmk_ well, I haven't yet even build tested about eight of these yet
16:00 matoking Oh, and it seems the mainline Linux kernel already has some Cubox-i-related changes in it
16:01 _rmk_ yes, that's me pushing it upstream, which provides basic support only atm
16:01 matoking Nice :)
16:02 _rmk_ I'll take a look at some point - but I'm going to be rather distracted at times because we've got quite a bit of storm damage from the last two months to deal with.
16:06 _rmk_ at the moment, it's still pretty windy so not going down the bottom of the garden in case of being squished by any further falling tree branches
16:37 jnettlet matoking, I found a bug with cpuidle that caused network latency problems. You should try disabling it and see if things look better.
16:48 matoking @jnettlet How do I do that?
16:49 jnettlet matoking, it will be under /sysfs/devices
16:50 jnettle 16:50 * jnettlet now questions if the 3.0.xx kernels had the sysfs entry to disable cpuidle
16:50 jnettlet do a find in /sysfs to see if you can find cpuidle. That kernel may be too old
16:50 matoking You mean /sysfs in root?
16:51 matoking Oh, right, sys
16:51 jnettlet yeah, it is sysfs mounted to /sys
16:52 jnettlet sorry doing too many things right now
16:52 mk01 # CONFIG_CPU_IDLE is not set
16:52 mk01 so there will be no /sys or debugfs entry
16:53 matoking Nope, can't find cpuidle under /sys
17:01 matoking Am I right in guessing that there are multiple kernel versions in use, since people on the forums are reporting better performance?
17:02 jnettle 17:02 * jnettlet has no idea
17:04 matoking Or they could be using the Ethernet port instead of Wi-Fi
18:54 warped-rudi jnettlet: I tried your debug u-boot, but no change. Behaviour exactly the same.
19:04 jnettlet warped-rudi, if you are bored you could install and run lmbench on both machines. I will run them on mine also so we can compare
19:17 warped-rudi I'm running geexbox on the machine. So I have to look how to cross-compile that tool...