| 00:37.43 | *** join/#nslu2-linux Jacmet (n=jacmet@50C59F0E.flatrate.dk) |
| 00:38.33 | dwery | Jacmet: hi. tested mtd and jffs2 on LE.. everything seems fine |
| 00:39.54 | [g2] | dwery so we're all set for now right ? |
| 00:40.22 | [g2] | re: perl performance (until I come up with something about PCI/AHB/USB/etc...) |
| 00:40.56 | dwery | [g2]: i copied the modules on the flash but the speed is almost the same. |
| 00:41.23 | dwery | [g2]: everything else seems fine, i think the currect ixp400 mtd code is perfect |
| 00:41.48 | [g2] | dwery you should try and copy the modules to a ramdisk |
| 00:42.20 | dwery | maybe i'll try, but i'll loose memory |
| 00:42.26 | [g2] | even copying perl to the ramdisk would be a good idea |
| 00:42.44 | [g2] | Oh... I'm just saying for a test case |
| 00:42.56 | [g2] | to see the CPU versus I/O timing |
| 00:43.03 | dwery | the system is running postgres, apache, courier and another bunch of programs :) |
| 00:43.18 | [g2] | you really need a loft :) |
| 00:43.26 | [g2] | 533 and 64M |
| 00:43.30 | dwery | probably yes ;0 |
| 00:44.01 | dwery | or use a cluster of NSLU2s :D |
| 00:44.12 | [g2] | that works too |
| 00:44.27 | dwery | can't remember if the Loft has onboard ide.. ? |
| 00:44.51 | [g2] | no but 2 MiniPCIs and 2 USB 2.0 plus a CF slot |
| 00:45.16 | dwery | how many spare I/O ? |
| 00:45.42 | [g2] | and with the latest Redboot you can do stuff like load ... -m disk hda1:/boot/zImage |
| 00:45.59 | [g2] | there's I2C, and 4 or 5 GPIOs |
| 00:46.14 | dwery | it would be fine to play with SPI and MMC |
| 00:46.26 | [g2] | why you have onbard CF |
| 00:46.44 | [g2] | I mean other than to just play |
| 00:46.56 | dwery | 'cos my camera (and most others) use MMC/SD |
| 00:47.11 | dwery | .. and to play, of course :) |
| 00:47.21 | [g2] | um... the USB 2.0 ports work great with an adapter |
| 00:47.27 | dwery | i hate adapters :) |
| 00:47.45 | dwery | jacmet was also thinking about adding an lcd to the slug |
| 00:48.06 | [g2] | I just saw the Logitech USB keyboard with built in LCD :) |
| 00:48.47 | dwery | WHERE? |
| 00:48.59 | dwery | (the logitech sitge, obviously :) ) |
| 00:49.57 | [g2] | http://www.extremetech.com/article2/0,1697,1870548,00.asp |
| 00:50.13 | [g2] | there |
| 00:50.31 | [g2] | there's an SDK for the LCD dunno if it's open or not |
| 00:50.51 | [g2] | I imagine the thing just shows up a multiple devices |
| 00:51.11 | [g2] | I've been very happy with the USB keyboard and mouse support in debian |
| 00:51.54 | [g2] | I've plugged in an extra keyboard and mouse on my thin client and I can use either one |
| 00:52.09 | [g2] | it's especially handy on the laptop for the mouse |
| 00:52.33 | [g2] | touch pad for a click or 1 point and grab the mouse for surfing |
| 00:53.14 | dwery | that keyboard is _huge_! |
| 00:53.19 | dwery | but very nice |
| 00:53.45 | [g2] | it's got the USB with built-in LCD and keyboard for 70-80 |
| 00:53.57 | *** join/#nslu2-linux Jacmet (n=jacmet@50C59F0E.flatrate.dk) |
| 00:54.09 | [g2] | it'd be cool to plug it in and just have the LCD |
| 00:54.18 | dwery | yep.. a bit expensive maybe |
| 00:54.40 | [g2] | like making you own is going to cost less |
| 00:54.57 | [g2] | 80 hours of time |
| 00:55.02 | dwery | no.. but there are usb lcds |
| 00:55.14 | [g2] | urls ? |
| 00:55.30 | dwery | mm.. can't remember, i think i got the link from the lcdproc website |
| 00:55.33 | [g2] | < $65 ? |
| 00:56.00 | dwery | proobably 70/80 USD for LCD+keyboard is better than 60 (or so) for LCD only.. |
| 00:57.09 | [g2] | depends on the LCD |
| 00:57.16 | [g2] | size, backlight etc.. |
| 00:57.56 | dwery | i'm searching for a red lcd with red backlight but no luck.. |
| 00:58.56 | dwery | given that the G15 can be found, according to google, for 60usd.. i think I might buy one.. |
| 00:59.21 | dwery | 55usd |
| 01:00.14 | dwery | but most of those shops will not ship to italy :( |
| 01:04.32 | *** join/#nslu2-linux nettie (i=esivieri@85-18-54-38.ip.fastwebnet.it) |
| 01:06.10 | [g2] | dwery we need to send a patch up stream that fixes all those cast warnings `__ixp4xx_writel' makes integer from pointer |
| 01:06.55 | dwery | [g2]: well, yes.. it's late in the night here, so i'm going to bed.. If you wanto to write this patch i could give it a test tomorrow :) |
| 01:07.57 | [g2] | I'm just saying in the near future |
| 01:08.08 | dwery | :D |
| 01:08.20 | [g2] | I'm testing the d-i installer rwhitby-away created |
| 01:09.19 | dwery | any major problem? |
| 01:10.59 | dwery | btw if you find any cheap usb lcd please drop me a note :) |
| 01:17.46 | [g2] | nod. |
| 01:18.46 | dwery | i also saw some usb "phones" with buil-it LCDs for ~50eur |
| 01:18.57 | dwery | s/buil-it/built-in/ |
| 01:20.40 | dwery | but the G15 is definitely better |
| 01:34.39 | dwery | i'm going, see ya.. |
| 02:38.51 | *** join/#nslu2-linux Jacmet (n=jacmet@50C59F0E.flatrate.dk) |
| 02:50.06 | *** join/#nslu2-linux [g2] (n=Tom@cpe-066-057-008-035.nc.res.rr.com) |
| 04:33.30 | *** join/#nslu2-linux ByronT (n=byron-po@nslu2-linux/ByronT) |
| 04:54.38 | *** join/#nslu2-linux CIA-8 (n=CIA@flapjack.navi.cx) |
| 05:57.34 | *** join/#nslu2-linux odoc (i=odoc@195.180.11.194) |
| 06:28.06 | *** join/#nslu2-linux a77ila (n=bear@mx.manufacture.co.yu) |
| 06:39.51 | *** join/#nslu2-linux caplink811-log (n=caplink8@dslb-084-059-086-234.pools.arcor-ip.net) |
| 07:30.02 | *** join/#nslu2-linux jelle (n=jelle@ip54518e8c.direct-adsl.nl) |
| 07:54.49 | CIA-8 | 03rwhitby * 10kernel/2.6.12/ (30-i2c-x1205.patch 50-nslu2-general.patch): Updated so that they compile against 2.6.12 |
| 07:57.08 | *** part/#nslu2-linux odoc (i=odoc@195.180.11.194) |
| 08:01.39 | rwhitby | dwery-away: nslu.sf.net kernel repo 2.6.12 ditr now compiles |
| 08:06.21 | *** join/#nslu2-linux odoc (i=odoc@195.180.11.194) |
| 08:28.57 | *** join/#nslu2-linux conradl (n=clancast@CPE-60-226-19-234.qld.bigpond.net.au) |
| 08:29.36 | CIA-8 | 03koen 07org.openembedded.dev * rbf441048... 10/conf/distro/preferred-gpe-versions-2.7.inc: preferred-gpe-versions-2.7.inc: update gpe-shield and gpe-bluetooth |
| 08:55.28 | rwhitby | conradl: we'e gonna get iomega nas100d boxen instead of ds101's. |
| 09:04.36 | CIA-8 | 03koen 07org.openembedded.dev * rc6c3b0d5... 10/packages/libgpewidget/libgpewidget_0.105.bb: libgpewidget: add 0.105 |
| 09:18.34 | conradl | rwhitby already looking for a good price! |
| 09:20.19 | rwhitby | conradl: USD$180 - you won't find it under $450 in aus |
| 09:20.36 | conradl | thats a challenge! |
| 09:21.01 | conradl | ordered a hx4700 today. |
| 09:22.44 | rwhitby | conradl: http://www.tigerdirect.com/applications/searchtools/item-details.asp?EdpNo=1118419&Sku=I21-2060 |
| 09:23.42 | dwery | rwhitby: hi. what's inside the 100d? |
| 09:24.47 | conradl | nice...but i want to try a via fanless m/b @667 in think. |
| 09:27.06 | rwhitby | dwery: armeb |
| 09:27.08 | rwhitby | have a look at www.debonaras.org / wiki / Info / BootLog |
| 09:27.11 | rwhitby | http://www.debonaras.org/wiki/Info/BootLog |
| 09:28.15 | conradl | looks like something we need to stock @ work... |
| 09:28.35 | rwhitby | http://nslu2-linux.thestuffguy.com/gallery/Iomega100d |
| 09:29.23 | rwhitby | twice the ram and flash of a slug, and wireless, and ide too. |
| 09:29.23 | dwery | seems a nice machine. around 450eur here |
| 09:29.42 | dwery | slow processor :( |
| 09:29.48 | rwhitby | yeah, seems that it's horrendously overpriced just about everywhere (including other places in the US). |
| 09:30.38 | rwhitby | but ka6sox has successfully ordered and received one from tigerdirect for USD$180 |
| 09:30.57 | dwery | tigerdirect won't sent outside us without a lot of troubles :( |
| 09:31.31 | dwery | do we know which rtc it uses? |
| 09:31.32 | rwhitby | that's why we have hacking communities to act as shipping agents ... |
| 09:31.38 | dwery | :-D |
| 09:32.49 | rwhitby | dunno - we don't have source code for it, just the firmware image (which is a simple tar file) |
| 09:32.59 | dwery | btw, I can confirm mtd works fine here. |
| 09:33.24 | dwery | iomega did not released the source? |
| 09:33.48 | rwhitby | cool - I finished getting the 2.6.12 version of your 2.6.14 patchset to compile today - it's checked in ready for someone else to test too :-) |
| 09:34.05 | dwery | rwhitby: i will. |
| 09:34.06 | rwhitby | apparently iomega released an unrelated package of source code. |
| 09:34.19 | dwery | the EFF would be happy to jump on them :) |
| 09:34.48 | rwhitby | ka6sox is trying the asking nicely approach first. |
| 09:35.33 | rwhitby | apart from the gpio settings. most other things on the board are known quantities for us, so it shouldn't be too hard to port 2.6.14 to it. |
| 09:36.34 | dwery | ack |
| 09:36.57 | rwhitby | dyoung and I have one each on the way ... |
| 09:38.33 | dwery | ehe :) i've to many boards on the desk right now.. |
| 09:39.57 | rwhitby | dyoung-away's gonna get his about a week before me, so he'll probably have all the fun |
| 09:41.12 | dwery | i;m on my second nslu2 and a third will be on the way sooner or later.. this game is not going to be cheap :) |
| 09:55.38 | jelle | usage of "leds" seems to be not as stated... |
| 09:55.56 | jelle | leds +0 => leds: 0: unknown LED (use r,g,0,1 or A) |
| 09:56.25 | jelle | should probably be "(use r,g,1,2 or A) |
| 10:26.52 | CIA-8 | 03rwhitby * 10debian/ (configs/nslu2_defconfig patches/linux-2.6.12-nslu2.patch): Updated to backported 2.6.14 patchset |
| 10:55.23 | CIA-8 | 03rwhitby * 10kernel/2.6.12/40-missing-exports.patch: Added the missing-exports patch from openembedded |
| 10:56.19 | CIA-8 | 03rwhitby * 10debian/patches/linux-2.6.12-nslu2.patch: Added the missing-exports patch from openembedded |
| 11:01.49 | CIA-8 | 03rwhitby * 10kernel/2.6.12/40-missing-exports.patch: Fixed the missing-exports patch |
| 11:02.03 | CIA-8 | 03rwhitby * 10debian/patches/linux-2.6.12-nslu2.patch: Fixed the missing-exports patch |
| 11:38.31 | *** join/#nslu2-linux jelle (n=jelle@ip54518e8c.direct-adsl.nl) |
| 11:44.21 | *** part/#nslu2-linux odoc (i=odoc@195.180.11.194) |
| 11:44.31 | *** join/#nslu2-linux odoc (i=odoc@195.180.11.194) |
| 11:49.34 | CIA-8 | 03koen 07org.openembedded.dev * rb7f6ccdd... 10/packages/libgpewidget/libgpewidget_cvs.bb: libgpewidget: update PV in the cvs .bb |
| 12:04.28 | *** join/#nslu2-linux cb_log (n=cbrake2@cpe-24-166-121-47.neo.res.rr.com) |
| 12:23.01 | *** join/#nslu2-linux twan_ (n=twan@18-129.242.81.adsl.skynet.be) |
| 13:03.25 | *** join/#nslu2-linux twan_ (n=twan@18-129.242.81.adsl.skynet.be) |
| 14:01.47 | *** part/#nslu2-linux odoc (i=odoc@195.180.11.194) |
| 14:15.10 | *** join/#nslu2-linux geojeff (n=geojeff@adsl-64-216-124-105.dsl.bumttx.swbell.net) |
| 14:21.14 | *** join/#nslu2-linux kolla_ (i=kolla@firda.uninett.no) |
| 14:22.30 | *** join/#nslu2-linux caplink811 (n=joerg@dslb-084-059-086-234.pools.arcor-ip.net) |
| 14:39.28 | VoodooZ_Log | [g2], have you measured the current consumption of the 533Mhz Loft? |
| 14:42.15 | CIA-8 | 03repvik * r220 10/releases/OpenSlug-2.7-beta/conf/distro/openslug-bbfiles.conf: Update openvpn from 2.0rc1 to 2.0. Fixes inline compiler bug |
| 14:42.16 | [g2] | VoodooZ_Log not yet |
| 14:42.38 | [g2] | but it shouldn't be too touch |
| 14:42.48 | [g2] | the 533 is cool enough to put my finger on |
| 14:42.51 | VoodooZ_Log | ok, I was just day-dreaming about one and how much faster my vision stuff would be. |
| 14:43.04 | NAiL | jelle: The feed is recompiling now, so it should be up in ~10 mins |
| 14:43.14 | VoodooZ_Log | the slug is already pretty hungry at .5 amp with all the extra junk on the board. |
| 14:43.43 | VoodooZ_Log | Then again, these things are not designed for battery operation so... |
| 14:43.50 | [g2] | nod |
| 14:44.09 | NAiL | Just get a big enough battery :-P |
| 14:44.29 | [g2] | I did notice we've got the Indirect mapping of the PCI turned on by default |
| 14:44.36 | VoodooZ_Log | hehehe. Not always possible unfortunately. more juice = more weight. |
| 14:44.56 | [g2] | I've been testing today without it |
| 14:45.59 | NAiL | What's the difference? |
| 14:46.09 | NAiL | I have no idea what indirect mapping is |
| 14:46.17 | [g2] | it's for the PCI mapping |
| 14:46.52 | NAiL | Any performance difference? |
| 14:48.29 | dwery-away | [g2]: I had noticed that too ans asked here but got no answer |
| 14:49.00 | dwery | given that the characteristics of the slug, direct mapping should b eok |
| 14:50.21 | [g2] | I don't have any hard data |
| 14:51.04 | [g2] | I've found the smoking gun with large files and thttpd |
| 14:51.32 | dwery | can you compare the performances? |
| 14:51.49 | [g2] | dwery before and after ? |
| 14:51.53 | dwery | yep |
| 14:52.02 | [g2] | it's really a little complicated |
| 14:52.20 | dwery | yeah. I don't know which kind of benchmark could be used |
| 14:52.34 | [g2] | meaning one has to disable most or all daemons and have a really similar setup |
| 14:53.00 | [g2] | with the daemons running suff moves around by a few MBs each run |
| 14:53.24 | [g2] | so that means either lots of runs or a specially crafted test setup |
| 14:56.19 | dwery | everything works fine with direct mapping? |
| 14:56.37 | [g2] | so far |
| 14:56.43 | dwery | ok, I'll switch too |
| 14:56.55 | [g2] | I haven't run into any issue yet |
| 14:57.24 | [g2] | dwery I was thinking I'd write up a patch to make it run time selectable if its is setup in an initialization area |
| 14:57.44 | dwery | mmmm |
| 14:58.03 | [g2] | then based on memory size one could set it up or not |
| 14:58.37 | dwery | for memory do you mean RAM? |
| 14:58.42 | [g2] | in the NSLU2 and Initial Lofts it'd be off, but some future lofts with >64MB might turn it on |
| 14:58.52 | [g2] | yes |
| 14:58.58 | dwery | I think we are talking about memory space in the PCI... |
| 14:59.26 | dwery | not system memory |
| 15:00.18 | [g2] | right it's the mapping of the PCI window size |
| 15:00.52 | [g2] | I don't know if that's per devices or bus or what |
| 15:02.29 | dwery | probably bus |
| 15:03.57 | [g2] | dwery I found the issue with thttpd and large files |
| 15:04.11 | [g2] | it mmaps the files up to around 1GB |
| 15:04.14 | dwery | how? |
| 15:04.17 | dwery | aaa |
| 15:04.21 | [g2] | so we wiind up swapping |
| 15:04.54 | [g2] | which is just totally crazy to me |
| 15:05.23 | [g2] | that reminds me I've got to go bug lennert about the swapping code |
| 15:44.14 | dwery | the only pci device on the slug is the usb controller and it's addressing requrtements are quite low |
| 15:44.20 | dwery | so i think direct addressing works just fine |
| 15:45.12 | dwery | it seems it doesn't support fast back to back transfers |
| 15:46.28 | dwery | [g2]: do you think there's space for optimizations? |
| 15:47.58 | *** join/#nslu2-linux Matta (n=matta@YYYDXL.dsl.saunalahti.fi) |
| 15:50.36 | CIA-8 | 03jbowler * 10upslug2/ (configure.ac nslu2_upgrade.cc nslu2_upgrade.h pcap_wire.cc): |
| 15:50.36 | CIA-8 | Improved pcap_wire.cc which uses getifaddrs |
| 15:50.38 | CIA-8 | Improved Wire error which includes the provision of a .what message |
| 15:50.40 | CIA-8 | configure.ac checks for getifaddrs |
| 16:05.17 | *** join/#nslu2-linux twan_ (n=twan@18-129.242.81.adsl.skynet.be) |
| 16:13.51 | *** join/#nslu2-linux eFfeM (n=slug@gtg-c-17077.adsl.wanadoo.nl) |
| 16:33.26 | *** join/#nslu2-linux odoc (i=odoc@195.180.11.194) |
| 17:47.10 | *** join/#nslu2-linux Tiersten (i=tman@nslu2-linux/Tiersten) |
| 17:50.06 | *** join/#nslu2-linux rwhitby-away (n=rwhitby@nslu2-linux/rwhitby) |
| 17:58.35 | *** join/#nslu2-linux AdamBaker (n=aab@userch028.dsl.pipex.com) |
| 18:01.51 | CIA-8 | 03jbowler * 10upslug2/ (ChangeLog upslug2.cc): |
| 18:01.52 | CIA-8 | Improved -t handling - -t is no longer required if there is only one |
| 18:01.52 | CIA-8 | NSLU2 in upgrade mode. |
| 18:01.52 | CIA-8 | Improved byte sex handling - new, more flexible, -e option (the old ones |
| 18:01.52 | CIA-8 | are gone) and automatic recognition of the kernel byte sex. |
| 18:01.52 | CIA-8 | MAC address filtering now accepts any MAC address. |
| 18:38.09 | *** join/#nslu2-linux mithro (n=tim@c213-100-42-188.swipnet.se) |
| 18:43.41 | [g2] | dwery back-to-back xfers was one of the things I wanted to try |
| 18:43.57 | [g2] | Also I think there's only 1 URB being used |
| 18:44.17 | dwery | [g2]: b2b on the slug or on the loft? |
| 18:45.01 | [g2] | If only one URB is really being used we could setup the Mini D-cache and statically map that for bulk transfers |
| 18:45.09 | [g2] | dwery both really |
| 18:45.52 | dwery | the usb controller on the slug does not support b2b |
| 18:46.01 | [g2] | There's a different USB controller on the Loft (Philips) |
| 18:47.01 | *** join/#nslu2-linux DaKa (n=david@193.13.58.211) |
| 18:47.10 | dwery | tell me about the mini d-cache |
| 18:49.49 | [g2] | there's a mini I cache and D cache on the chip each 2K |
| 18:49.58 | [g2] | they run at core speed |
| 18:50.08 | dwery | ok |
| 18:50.15 | [g2] | so that'd be 266 for NSLU2 turbo and 533 for Loft |
| 18:50.31 | [g2] | which are 2x and 4x faster than the memory @ 133 |
| 18:51.41 | [g2] | So I'd imagine if we could reservere 4 512 byte blocks for the data areas we could support 4 devices or 2 devices with two transactions outstanding |
| 18:51.42 | dwery | the 1-urb problem is at usb-storage or usb-core level? |
| 18:52.16 | [g2] | I'd think it be in the usb-storage or usb-core |
| 18:52.50 | dwery | so we should map the memory area used by the urb on the cache, right? |
| 18:53.16 | [g2] | you can dump some usb info and only 1 transfer is allowed for the bulk device |
| 18:53.35 | [g2] | dwery that's exactly my though about the mapping |
| 18:53.55 | dwery | how that could be done? I gues you can't malloc() :-D |
| 18:54.12 | [g2] | I don't know how easy it is, worst case it's custom patches to the ehci/usb drivers |
| 18:54.42 | [g2] | one needs to look at the URB allocation and I think we'd have to map some there |
| 18:54.45 | dwery | where's that cache mapped? |
| 18:55.26 | [g2] | right now I'm not sure if it's used or how it gets mapped |
| 18:55.55 | [g2] | this is part of the beauty of Open Source and open hardware, we can really tune this stuff |
| 18:56.10 | dwery | ok.. keep me informed.. if you have some spare time in the next few days we can examine it |
| 18:56.30 | [g2] | nod I'm trying to get a handle on it |
| 18:56.52 | [g2] | clearly, several minds are better than 1 |
| 18:57.34 | [g2] | so you've checked that the NECdoesn't support back-to-back xfers ? |
| 18:57.58 | dwery | i used lspci -v |
| 18:58.05 | dwery | and the fast flag is not present |
| 18:59.56 | *** join/#nslu2-linux philipjohnson (n=philipjo@host86-130-143-196.range86-130.btcentralplus.com) |
| 19:04.08 | mithro | hey people |
| 19:04.17 | mithro | [g2]: you comming to Linux.conf.au? |
| 19:05.18 | *** part/#nslu2-linux odoc (i=odoc@195.180.11.194) |
| 19:05.31 | *** join/#nslu2-linux odoc (i=odoc@195.180.11.194) |
| 19:07.57 | [g2] | mithro I'd like to go can you fly me there ? :) |
| 19:09.44 | mithro | [g2]: tax deduction? |
| 19:10.00 | mithro | you could pimp your arm device |
| 19:10.07 | [g2] | hey I've got enough of those |
| 19:10.14 | [g2] | I need cash man :) |
| 19:10.30 | [g2] | or at least travel |
| 19:10.42 | [g2] | I'll do the local circuit first |
| 19:12.52 | mithro | btw where are you located? |
| 19:13.18 | [g2] | near Raleigh NC |
| 19:13.28 | mithro | in the USA? |
| 19:13.31 | [g2] | yep |
| 19:13.53 | [g2] | what part of the world to you hail from ? |
| 19:14.12 | mithro | i'm from Australia |
| 19:14.18 | mithro | but currently residing in Sweden |
| 19:19.57 | philipjohnson | Does anyone know about the nslu2-linux.org downtime? |
| 19:20.18 | philipjohnson | Well, why it's down? |
| 19:21.17 | *** join/#nslu2-linux jelle (n=jelle@ip54518e8c.direct-adsl.nl) |
| 19:22.02 | senneth | someone broke the internet |
| 19:22.33 | jelle | good thing i have internet on my computer... |
| 19:22.37 | jelle | lol |
| 19:22.39 | philipjohnson | :D |
| 19:39.14 | [g2] | dwery-away where do you see the back-to-back on the lspci -v ? |
| 19:39.49 | dwery-away | i do NOT see it |
| 19:39.56 | dwery-away | while on my pc i see the fast flag |
| 19:40.47 | dwery-away | also in dmesg: PCI: bus0: Fast back to back transfers disabled |
| 19:41.08 | dwery-away | the b2b feature gets disable if there's at least one device which does not support it |
| 19:41.21 | dwery-away | arm/kernel/bios32.c (iirc) |
| 19:41.48 | [g2] | well the Philips part can be a bus master |
| 19:42.11 | [g2] | I haven't tried setting it up that way, but it's got some capabilities |
| 19:42.22 | [g2] | thx for the info |
| 19:42.28 | [g2] | I'lll check some more later one |
| 19:42.32 | [g2] | on |
| 19:42.47 | [g2] | bbl |
| 19:42.54 | dwery-away | maybe i'm wrong, still have to study the problem. you may want to try an lspci -v and -vv on the loft |
| 19:43.06 | [g2] | that's what I was looking at :) |
| 19:43.22 | dwery-away | i'll be back later in the night |
| 19:43.31 | [g2] | ah... FastB2B- |
| 19:49.39 | *** join/#nslu2-linux yvasilev (n=yvasilev@201.135.158.15) |
| 20:08.41 | *** join/#nslu2-linux kolla_ (i=kolla@firda.kolla.no) |
| 20:11.49 | *** part/#nslu2-linux odoc (i=odoc@195.180.11.194) |
| 20:51.35 | *** join/#nslu2-linux jelle (n=jelle@ip54518e8c.direct-adsl.nl) |
| 20:55.00 | *** join/#nslu2-linux odoc (i=odoc@195.180.11.194) |
| 21:38.01 | rwhitby-away | [g2]: cvs 2.6.12 patchset is now a back-ported version of the collaborative 2.6.14 patchset |
| 21:56.14 | [g2] | rwhitby thx |
| 22:01.36 | kolla_ | mf.. buhu, ipv6 is still not on per default |
| 22:01.49 | kolla_ | now, which of all this kernel configs to edit... |
| 22:01.53 | kolla_ | these |
| 22:04.12 | kolla_ | hm, but it has ipv6 there, strange |
| 22:06.01 | kolla_ | ah, it is there :) |
| 22:20.33 | dwery | [g2]: I gave a quick look at the usb-storage code. only one urb it seems. |
| 22:21.02 | rwhitby | dwery: hi |
| 22:21.14 | dwery | rwhitby: hi! |
| 22:21.38 | rwhitby | dwery: had a chance to test the 2.6.12 backport? |
| 22:21.54 | dwery | rwhitby: no, both slugs are busy |
| 22:22.10 | dwery | i should be able to free the development one ina couple of days |
| 22:23.47 | dwery | I emailed deepak about the patch's status, awaiting answer. |
| 22:24.14 | rwhitby | ok |
| 22:24.47 | rwhitby | can you do a diff between the 2.6.12 patchset and the 2.6.14 patchset and eyeball review the small set of changes? |
| 22:25.12 | dwery | yep, just a moment |
| 22:30.41 | dwery | linux-2.6.12-nslu2.patch is confusing me |
| 22:30.50 | dwery | it seems there's an x1205 driver |
| 22:31.32 | dwery | plus a lot of other things? duplicate? |
| 22:32.00 | rwhitby | dwery: ignore that file - that is the OE patchset which has been replaced |
| 22:32.23 | rwhitby | (i.e. that was the previous 2.6.12 patchset before your work started) |
| 22:32.30 | dwery | others are ok, at a first glance. |
| 22:33.36 | rwhitby | thx |
| 22:34.36 | dwery | rwhitby: did you read the thoughts about the usb storage urbs and d cache? (g2) |
| 22:35.55 | rwhitby | dwery: yeah, saw that. interesting stufff - far too low level to me :-) |
| 22:36.22 | dwery | do we have someone with experience on this stuff? |
| 22:43.58 | rwhitby | dunno |
| 22:51.06 | *** join/#nslu2-linux yvasilev (n=yvasilev@201.135.158.15) |
| 22:56.56 | dwery | ok, the mini data cache cannot be used as data ram |
| 22:57.10 | dwery | developer's manual, page 67 |
| 22:59.49 | dwery | we could reserve a portion of the 32k cache, but it's not easy. |
| 23:04.48 | dwery | i don't know if anyone ever tried this on linux. |
| 23:16.01 | [g2] | dwery I thought I read it could be used as static memory for a device driver |
| 23:16.33 | [g2] | certainly stealing some of the d-cache would work too |
| 23:16.39 | dwery | you could use the 32k cache and split it, like 28K data cache and 4 k data ram, but not the mini i cache, if i've read correctly the manual. |
| 23:16.44 | dwery | s/mini i/mini d/ |
| 23:16.56 | rwhitby | back later |
| 23:17.38 | [g2] | it appears a few Coyote boards have the fast back-to-back xfers enabled |
| 23:19.02 | dwery | the data cache thins is pretty hardcoded in the assembly files |
| 23:19.03 | [g2] | dwery does the usb-storage only use 1 URB ? |
| 23:19.14 | dwery | it seems so, if i've understood the code |
| 23:19.26 | [g2] | dwery that's what machine specific code is for right ? |
| 23:19.37 | dwery | yes |
| 23:19.40 | [g2] | ::) |
| 23:20.20 | [g2] | so there's really only 1 URB used by the bulk devices ? |
| 23:20.57 | dwery | i think so |
| 23:21.07 | dwery | and I also think the 2k mini d cache is not used at all |
| 23:21.10 | *** part/#nslu2-linux odoc (i=odoc@195.180.11.194) |
| 23:21.11 | [g2] | that's just semi-crazy to me |
| 23:21.40 | [g2] | nod. on the mini cache -- we should put that to use for _something_ :) |
| 23:21.42 | dwery | or maybe yes. i'm confused :) |
| 23:21.56 | dwery | you can use it as a cachem but not as data ram |
| 23:22.04 | dwery | now don't ask me the difference :D |
| 23:22.55 | [g2] | well from the kernel drivers point-of-view I don't know if it much matters |
| 23:23.41 | [g2] | from my very limited understanding there are different allocators in the kernel with virtual memory being used mostly |
| 23:24.00 | [g2] | only a couple places deal with physical memory |
| 23:24.09 | [g2] | like the virtual memory allocator |
| 23:24.44 | dwery | yep. but configuring the cache is not exactly easy. |
| 23:25.05 | dwery | especially when you are going to setup a 28K cache in place of a 32 one :) |
| 23:25.15 | dwery | we'd need the people who originally coded the cache thingy |
| 23:25.48 | [g2] | I wouldn't want to steal 4K |
| 23:26.00 | [g2] | I'd steal like 512 bytes |
| 23:26.31 | [g2] | is that 8 or 16 cache lines ? |
| 23:26.36 | dwery | i guess the problem is not how much you are stealing, but the fact that you are stealing :) |
| 23:26.58 | dwery | a line is 32 byte |
| 23:27.10 | [g2] | so 16 |
| 23:27.12 | dwery | yes |
| 23:27.59 | [g2] | I'd be willing to bet that disk performace would significantly improve |
| 23:28.16 | [g2] | especially if we took 2 |
| 23:28.20 | dwery | i'm not so sure |
| 23:28.36 | dwery | you have to send the urb down the pci to the usb controller |
| 23:29.47 | dwery | and I think that is the part which requires most of the time |
| 23:30.22 | [g2] | The PCI runs at 33Mhz 4 bytes wide right ? |
| 23:30.50 | dwery | i think so. there's an option for 66, but it's not used on the slug i think |
| 23:30.52 | [g2] | the AHP runs 133 with bursts 32 bytes long or 8 PCI writes |
| 23:31.28 | [g2] | so the PCI is off the South side AHP |
| 23:31.32 | dwery | yes |
| 23:32.13 | [g2] | so from the XScale to the PCI 32 bytes could get there in 2 33Mhz clock ticks right . |
| 23:33.47 | [g2] | the "slow" part is the PCI bus to the USB controller but we should have DMA there |
| 23:34.39 | dwery | is dma actually used? |
| 23:35.23 | [g2] | I sure hope so |
| 23:35.53 | [g2] | The 533 does 14+ MBs to my Maxtor drive |
| 23:36.08 | [g2] | I hope it's not do all that 4 bytes at a time |
| 23:36.19 | [g2] | that processor that is |
| 23:36.55 | [g2] | I see sustained xfers > 8MBs and the processor is 20% loaded ant 80% wait |
| 23:37.26 | [g2] | so I'm guessing it's in a tight loop waiting for the dma to complete |
| 23:37.51 | [g2] | I think the processor is avialable for other stuff |
| 23:38.28 | dwery | did you tried the same drive on an x86? |
| 23:38.44 | [g2] | but of course! |
| 23:38.55 | [g2] | :) 24 or 28MBs |
| 23:39.15 | [g2] | that's on the 2.4 P4 (I haven't tried on the AMD64) |
| 23:39.56 | [g2] | so I'm only getting 50-66% of the acess of the drive with the 533 |
| 23:40.11 | dwery | ok |
| 23:40.26 | [g2] | I get 8.5 on the Ultra II CF which tops out at 10MBs on the P4 |
| 23:40.39 | [g2] | that's 85% native speed |
| 23:41.06 | dwery | the nec usb controller is bus mastering capable. do you know if that;s used? |
| 23:43.20 | [g2] | the pullups on the EXP bus determine whether the IXP is Master or Target |
| 23:43.30 | [g2] | I think it's setup as Master |
| 23:43.42 | [g2] | the dmesg says ... PCI.... is Host |
| 23:44.16 | [g2] | the was an issue with this on the Loft with CF initially |
| 23:44.36 | [g2] | a Microdrive was pulling that line low and the PCI was coming up Target |
| 23:46.11 | dwery | i don't know very well the pci bus, but I remember that on the Amiga and its zorro bus the boards were able to exchange role.. but we also had a memory controller chip |
| 23:47.42 | [g2] | IXP42x too. the pullup determines it |
| 23:48.18 | [g2] | I think the Philips chip can run at 66Mhz too |
| 23:48.33 | [g2] | I can cut a resistor an possibly run at 66Mhz |
| 23:48.53 | dwery | so the usb controller can pull down and write to the memory, right? |
| 23:49.16 | [g2] | but then the MiniPCIs wouldn't work right |
| 23:49.43 | [g2] | I think either target or master can do DMA |
| 23:50.05 | [g2] | it just the Master is the bus arbiter and controller of the clk |
| 23:51.08 | dwery | ok |
| 23:51.13 | dwery | what about latency on the pci? |
| 23:55.27 | [g2] | from my understanding there's latency, burst size, back-to-back xfers , etc... |
| 23:56.09 | [g2] | we may be seeing the limits of a single transaction |