**** BEGIN LOGGING AT Mon Jul 18 02:59:57 2011
Jul 18 06:16:58 <hell> morning
Jul 18 06:17:08 <hell> can i get whole mtd as device?
Jul 18 06:23:35 <philipp64|laptop> Weedy: is that fast enough?
Jul 18 06:38:07 <CIA-89> nbd * r27680 /trunk/package/uci/Makefile: uci: update to 2011-07-18, fixes #9731
Jul 18 09:33:10 <WillieNL> philipp64|laptop: for raid controllers double check on test reports, for example disk fails and you can't order a disk with the same size ;-) 3ware will not like that :p
Jul 18 09:33:37 <WillieNL> areca's are more flexible about that..
Jul 18 09:35:03 <WillieNL> personal i'll go areca or onboard chipset.. everything else dunno just a waste when it is only for a little bit more performance.
Jul 18 09:36:27 * WillieNL is still happy with the md raid performance results
Jul 18 09:37:03 <WillieNL> atleast afther some tests i did not even want a raid controller anymore.
Jul 18 09:37:11 <roh> use a marvell controller and do linux softraid
Jul 18 09:37:54 <roh> add a decent amount of ram as cache
Jul 18 09:40:44 <EqUaTe> s/marvell/intel/
Jul 18 09:42:54 <roh> doesnt matter. only 'proper not broken sata controller' ;)
Jul 18 09:44:40 <WillieNL> roh: hmm a 3 disk raid5 on pata-133 disk's was performing spectaculous back in the day compared to lsi(slow) and 3ware(generic)
Jul 18 09:45:03 <WillieNL> so md raid is not bad at all when you know wat your doing
Jul 18 09:45:12 <WillieNL> and wat you want/need
Jul 18 09:45:17 <roh> WillieNL: its not a matter of the ata link. in the end all caches on the disks are disabeld anyhow when you do sw raid
Jul 18 09:45:52 <roh> i only suggested marvel because i know they have add-on pci-e cards with 8 sata ports for fair money
Jul 18 09:46:03 <WillieNL> roh: also but when you have used those products in production and have been ranting about issues you will not pay that price a 3th time again
Jul 18 09:46:38 <WillieNL> like 3 ware and forced todo a lot more work then just swap a disk.. stuff like that can really suck when it is not a hobby
Jul 18 09:46:57 <WillieNL> then i will say -> md raid is fine
Jul 18 09:47:08 <roh> md raid isnt bad, no. on the contrary. you have much better management and long-term stability features. you can be sure you can continue using it with all hw which comes along, since the on-disk-layout is known and done by opensource.
Jul 18 09:47:24 <WillieNL> indeed
Jul 18 09:47:38 <WillieNL> sadly only those areca controllers did make me really happy..
Jul 18 09:48:25 <WillieNL> roh: but ehm disk cache get disabled when use md raid ? thats new for me..
Jul 18 09:48:29 <roh> i've used a lot stuff the last 10 years and must say i dont see a reason for hw raid. if its to slow you have too slow hw. (usually the disks) or a workload which needs more caching.
Jul 18 09:50:00 <roh> WillieNL: you do not WANT caches on the disks anyhow. you want to know its written when it says so. caches dont help you there (the sata link is so much faster than the disk itself). also its p2p, so no other devices which could use the bandwith when not used by that one disk. you will need a block cache, sure. but thats on the /dev/mdX level then and not on the /dev/sdX level
Jul 18 09:50:24 <WillieNL> hmm example via based desktop board with 3 pata disk`s and have a result like +-250~300M/s write and 450M/s read performance ..
Jul 18 09:50:26 <roh> means the 'cache between fs and disk' moves from the disk to the main ram
Jul 18 09:50:48 <WillieNL> i calculated a little bit cache in there afther have seen the results from dd moving a few gigs
Jul 18 09:51:05 <WillieNL> while those disk`s normal can't do better them 50M/s each
Jul 18 09:51:11 <roh> imho atom and via hw is bad. atleast for throughput. use something with decent cpu caches and io bandwith
Jul 18 09:51:21 <WillieNL> this was on a via chipset :)
Jul 18 09:51:31 <WillieNL> and never tested my intel boards the same way btw :(
Jul 18 09:52:12 <roh> if you dont need to saturate gbit links with your raid basically anything will work which accepts >2 disks
Jul 18 09:52:58 <WillieNL> hmm jup or use modern sata drives.
Jul 18 09:53:47 <roh> btw.. the 'thecus' nas hardware is a p3 celeron based intel chipset board (custom) with marvel sata links and up to i think 7 internal slots
Jul 18 09:53:55 <roh> atleast the 'pro' and up versions
Jul 18 09:54:06 <WillieNL> heh 440bx based p3 ?
Jul 18 09:54:24 <WillieNL> hmm no they had no sata in that time .
Jul 18 09:54:25 <roh> nah. much newer. embedded p3. not that 'hot'
Jul 18 09:54:39 <WillieNL> hmm i bumped into a centrino desktop bord 2 weeks ago..
Jul 18 09:54:58 <roh> the sata is done with the marvell chip (8 lanes, 5 internal ones, one external one in the 5 slot enclosure)
Jul 18 09:55:17 <WillieNL> AOpen i855GMEm-LFS.. mobile centrino cpu supported.
Jul 18 09:55:46 <roh> eats about 80W idle with 5 disks and i think <40 if the disks are spun down
Jul 18 09:55:53 <WillieNL> still looking around for a bracket moving the 2 sata connectors from internal to a unused pci slot bracket.
Jul 18 09:56:10 <WillieNL> hmm this runs with 2.5" disk at 19.8W :)
Jul 18 09:56:54 <WillieNL> anyway.. gotta start the day here.. short night..
Jul 18 09:57:00 <roh> hf
Jul 18 09:57:58 <WillieNL> roh: hmm anyway if it is socket 370 i still have a p3 700 somewhere around.
Jul 18 09:58:07 <WillieNL> lv p3
Jul 18 09:58:18 <WillieNL> sl4ch
Jul 18 09:58:41 <WillieNL> hmm lv->coppermine
Jul 18 10:06:58 <Weedy> philipp64|laptop: when i'm in the shitty money mood i buy areca
Jul 18 10:15:16 <CIA-89> nbd * r27681 /trunk/package/uci/Makefile: uci: update to fix more unused-but-set issues (#9731)
Jul 18 10:37:10 <Weedy> philipp64|laptop: it depends on the board really, what you need is that sata subsystem to be a seperate as possible from all the other shit
Jul 18 10:38:15 <Weedy> if your board has really crappy throuput you can get cheap addin card and choose pci(-e) slots that are isolated to get the throughput
Jul 18 10:38:18 <CIA-89> swalker * r27682 /packages/ipv6/radvd/ (Makefile files/radvd.init): [packages] radvd: add new 1.8 options, split radvdump into a separate package
Jul 18 10:38:38 <Weedy> they thing is software raid need more bandwidth the hardware raid
Jul 18 10:40:00 <jow_laptop> swalker: are you sure its "Base6Interface" ?
Jul 18 10:40:16 <jow_laptop> and not Base6to4Interface
Jul 18 10:41:54 <swalker> jow_laptop: yes, 26/02/2011 Added Base6Interface option
Jul 18 10:42:01 <jow_laptop> oh ok
Jul 18 10:42:05 * jow_laptop is out of the loop
Jul 18 10:43:22 <swalker> 1.7 -> 1.8 diff, http://pastebin.com/8bMxsnLj
Jul 18 10:43:58 <jow_laptop> 90% autoconf noise :P
Jul 18 10:44:26 <swalker> correct :/
Jul 18 10:45:13 <jow_laptop> the Base6Interface idea looks nice
Jul 18 10:45:47 <jow_laptop> so you can basically build a static config for a dynamically changing upstream interface
Jul 18 11:00:56 <CIA-89> swalker * r27683 /packages/ipv6/radvd/patches/: [packages] radvd: remove empty patches directory
Jul 18 11:04:19 <WillieNL> Weedy: jup true sw raid do need a bit more bandwith but a reneric amd k7 will do compared to a hw raid card with onboard cpu..
Jul 18 11:04:43 <WillieNL> Weedy: simply nowdays server mainboards will not be any problem.. or they must be really crappy indeed..
Jul 18 11:05:52 <WillieNL> *argh* open the front door en it starts raining again :S
Jul 18 11:34:10 <seuros> Who is working on danube platform ?
Jul 18 11:42:47 <CIA-89> swalker * r27684 /packages/ (9 files in 9 dirs): [packages] remove empty patches directories
Jul 18 11:44:32 <tripolar> hi
Jul 18 11:45:09 <tripolar> what do  a have to add to a CMakeLists.txt file to get the executeable linked against libpthread?
Jul 18 11:49:50 <nbd> use TARGET_LINK_LIBRARIES
Jul 18 11:50:06 <nbd> first argument is the name of the target that you want to add the library dependency for
Jul 18 11:50:13 <nbd> following arguments specify the libraries to link against
Jul 18 11:52:22 <tripolar> thanks nbd
Jul 18 12:50:17 <CIA-89> jow * r27685 /branches/packages_10.03.1/ipv6/radvd/ (Makefile files/radvd.init patches/): [packages_10.03.1] merge r27682, r27683
Jul 18 14:23:55 <CIA-89> jow * r27686 /trunk/package/uhttpd/ (6 files in 2 dirs): [package] uhttpd: support building against openssl instead of cyassl, minor cleanups (#7827)
Jul 18 15:56:13 <blogic> luka12346|wiik: ping
Jul 18 16:32:36 <CIA-89> blogic * r27687 /trunk/target/linux/lantiq/patches-2.6.39/999-fix-eiu.patch:
Jul 18 16:32:36 <CIA-89> [lantiq]
Jul 18 16:32:36 <CIA-89> * fix external interrupts
Jul 18 17:01:24 <philipp64|laptop> did mtd break again?  I built yesterday morning (Pacific time) and reloaded, but I'm seeing the "failed to whiteout" messages again. I thought that got fixed?
Jul 18 19:16:17 <luka12346|wiik> blogic: pong
Jul 18 19:57:40 <blogic> luka12345|wiik: hi
Jul 18 19:57:49 <blogic> luka12345|wiik: how does the keys thing work in userland ?
Jul 18 20:01:59 <luka12345|wiik> blogic: you can use event_test package: 'event_test /dev/input/event0'
Jul 18 20:02:06 <luka12345|wiik> that is used for testing
Jul 18 20:02:58 <luka12345|wiik> and for advanced usage, ie. execute script when button is pressed i'm using triggerhappy
Jul 18 20:04:00 <blogic> ok
Jul 18 20:04:03 <luka12345|wiik> there is also package cmdpad, but i did not use it
Jul 18 20:05:35 <luka12345|wiik> dont forget to compile kernel module kmod-input-gpio-keys-polled; otherwise your button wont get recognized
Jul 18 20:09:17 <blogic> yes
Jul 18 20:09:28 <blogic> i will add the buttons to all targets next few days
Jul 18 20:09:43 <luka12345|wiik> great :)
Jul 18 20:10:11 <luka12345|wiik> do you maybe remember what is other button gpio for sx763?
Jul 18 20:10:44 <blogic> nope
Jul 18 20:10:48 <luka12345|wiik> ok
Jul 18 20:10:49 <blogic> i cant find the piece of paper
Jul 18 20:12:24 <luka12345|wiik> in '/sys/class/gpio/' i see 'gpio29 gpiochip0 gpiochip16 gpiochip48'
Jul 18 20:12:48 <luka12345|wiik> i did not define that values anywhere...
Jul 18 20:13:36 <blogic> gpiochip0 gpiochip16 gpiochip48 are the 3 gpio blocks
Jul 18 20:13:38 <blogic> gpio29 gpiochip0 gpiochip16 gpiochip48'
Jul 18 20:13:44 <blogic> grr
Jul 18 20:14:10 <blogic> gpio29 is GIGASX76X_USB
Jul 18 20:14:20 <luka12345|wiik> ah, yes
Jul 18 20:16:23 <luka12345|wiik> showld we put dev-gpio-buttons.(c|h) on one place in kernel? i have copied it from ar71xx and i guess it's also in other places...
Jul 18 20:17:31 <blogic> not sure
Jul 18 20:20:40 <blogic> i'll put on the list
Jul 18 20:41:26 <alesan> so I am compling OpenWRT with make -j 16 but there is only ONE CPU being used 100%
Jul 18 20:41:38 <alesan> all the others are almost not used
Jul 18 20:41:46 <alesan> is there a BUG?
Jul 18 20:51:02 <alesan> even compiling the Linux kernel is done with only one CPU
Jul 19 00:09:07 <mazilo> Does OpenWRT support packages written in a C# language?
**** ENDING LOGGING AT Tue Jul 19 02:59:56 2011