**** BEGIN LOGGING AT Mon Jan 22 02:59:58 2007 Jan 22 03:28:26 iwo: I already have a package setup for ov511-2.32 if you want it. Jan 22 03:28:54 Just ping me next time you're on. Jan 22 03:30:56 VoodooZ: how come it's not in OE yet? Jan 22 03:35:15 03osas * r5257 10optware/trunk/make/iksemel.mk: iksemel: iksemel-1.2 Jan 22 03:37:55 03osas * r5258 10optware/trunk/Makefile: iksemel: iksemel-1.2 ready for testing Jan 22 08:13:54 Hi guys Jan 22 08:14:56 I am running a unslung @266MHz with kernel 2.4.22 compiled with current sources Jan 22 08:15:29 I am wodering about POOR, really poor network performance both in NFS and samba Jan 22 08:15:54 255MB take 17minutes to copy using nfs Jan 22 08:16:47 I tried to track down the bottleneck and I am almost sure its the network Jan 22 08:17:04 the box is directly connected by crosscable Jan 22 08:17:22 It can't be the USB bus Jan 22 08:17:40 tried different cables? The slug appears to be very sensitive wrt network cabling. Jan 22 08:17:46 I connected the source-disk directly and copying is MUCH faster Jan 22 08:18:16 ok, wait - I take 100MBit switch and 2 regular cables Jan 22 08:21:43 ruschi: compared against the performance figures on the wiki? Jan 22 08:25:10 I can't find the performance figures in the wiki Jan 22 08:25:22 255MB in 17 minutes is about 250KB/sec. Running a deunderclocked slug, I'd expect ~5mb/sec from samba. Jan 22 08:25:46 When I tune everything I can, I get about 7MB/sec from samba on openslug. Jan 22 08:25:59 ok, I found it Jan 22 08:26:14 yes - I got about 150kB/s Jan 22 08:26:32 with nfs volume mounted rw,sync Jan 22 08:26:53 yeah there are 250kBit/s Jan 22 08:27:48 "iptraf" on the client gives me a TCP-Flow rate of 150KBit/s Jan 22 08:28:24 the loadavg on the slug is : 1.73 1.51 1.33 2/31 299 Jan 22 08:28:52 there are about 30MB of RaM used on the slug Jan 22 08:29:19 any idea of the bottleneck? Jan 22 08:29:59 The loadavg looks ok (the network driver always adds 1 to the loadavg) Jan 22 08:30:26 You've tried with both a direct connection and through a switch? Jan 22 08:31:18 if I look at /proc/interrrupts I have ahigh interrupt count with IXP425 Timer 1, ixp425_eth and ehci_hcd -just as expected Jan 22 08:31:45 yes currently testing with a 100MBit switch and patch cables Jan 22 08:33:18 using wireshark (a.k.a ethereal) ther are no anomanlies like TCP window congestion or retransmits either.... Jan 22 08:35:41 which package for unslung contains top? Jan 22 08:38:53 again - same result with swich..... Jan 22 08:39:24 ruschi: found top yet? Jan 22 08:39:39 16:59 minutes for 255MB Jan 22 08:39:54 no - the package search on nlu2-linux.org is really slow Jan 22 08:40:00 it's procps Jan 22 08:40:27 thanks Jan 22 08:51:19 wait - reading is realy fast: the same file reading wiht nfs gives 2:58 for 255MB Jan 22 08:56:37 "sync" will slow down your writes dramatically Jan 22 08:56:58 so mount the stuff without sync? Jan 22 08:57:16 isn't this dangerous for data integrity? Jan 22 08:57:33 can't have both speed and safety :-) Jan 22 08:57:52 are samba shares mounted synced? Jan 22 08:58:05 I only know the sync option form NFS Jan 22 09:06:10 1000 THANKS TO rwhitby: mounting the share with async gives the ultimate performance-boost! Jan 22 09:06:40 the time is now down to 1 minute!! Jan 22 09:10:07 OK see you later I do now teset using windows XP ans samba Jan 22 10:17:49 03oleo * r5259 10optware/trunk/make/transmission.mk: transmision: r1424->r1425 Jan 22 11:28:44 03oleo * r5260 10optware/trunk/make/vnstat.mk: vnstat: change cron.d/vnstat filemode Jan 22 12:21:14 03operron * r5261 10optware/trunk/make/ds101-kernel-modules.mk: ds101-kernel-modules.mk: bump up IPK_VERSION + minor fix in prerm generation Jan 22 13:34:39 rwhitby: re: OV511 package. Well, the reason I kept it for myself is that the kernel already includes an older subset of it. Jan 22 13:35:15 I do have the latest version as a .bb file but I wasn't sure if it was worth pushing it. Jan 22 13:35:27 I can send you the file if you want to add it. Jan 22 15:37:11 03bzhou * r5262 10optware/trunk/ (Makefile make/iksemel.mk): promoted iksemel with small fix Jan 22 15:38:57 morning Jan 22 15:39:22 morning Jan 22 16:25:38 03gda * r5263 10optware/trunk/make/sqlite.mk: sqlite: 3.3.9 -> 3.3.11 Jan 22 17:11:53 03oleo * r5264 10optware/trunk/ (4 files in 3 dirs): eletric-fence: C memory debugging library - added Jan 22 17:50:28 eno: thx for the iksemel fix :-) Jan 22 18:21:51 03osas * r5265 10optware/trunk/make/asterisk14.mk: asterisk14: now with jabber support Jan 22 20:35:09 does debian on the slug set up a serial console? Jan 22 20:43:45 looks like there is :-) Jan 22 20:48:51 cbrake: yes it does. 115200,8n1 Jan 22 20:49:04 (all our slug kernels do) Jan 22 20:52:01 rwhitby: excellent Jan 22 20:52:49 rwhitby: for some reason when I plugged a serial cable into it all I got was garbage -- even at 115200 Jan 22 20:52:57 rwhitby: then I rebooted and serial console was fine. Jan 22 20:54:52 probably out of sync Jan 22 22:28:21 folks, i want to hot-swap between dropbear and openssh on a running unslung 5.5 host in order to access it via sftp Jan 22 22:28:38 i need to ipkg remove dropbear before i can ipkg add openssh? Jan 22 22:29:20 the openssh-sftp package mentioned in the wiki has been rolled into the main openssh package? Jan 22 23:35:14 03g2-tbillman * r673 10kernel/trunk/patches/2.6.19/ (5 files): (log message trimmed) Jan 22 23:35:14 Fix 2.6.19 Loft eth1 oops on 0.3.1 firmware due to incorrect Jan 22 23:35:14 rxdoneq_id setup. THX to Christian for helping to identify Jan 22 23:35:14 the issue. THX to Gordon for the initial backport. Jan 22 23:35:14 The testing loft_defconfig is included. Jan 22 23:35:14 Eth0 worked, but that was more luck than proper initialization. Jan 22 23:35:19 The rxdoneq_id had a value of 0 instead of 4 as in ixdp425-setup.c. Jan 22 23:51:07 03g2-tbillman * r674 10kernel/trunk/patches/2.6.20/ (4 files): Jan 22 23:51:07 Proper rxdoneq_id setup for GPL 0.3.1 firmware on several Jan 22 23:51:07 targets. Jan 23 00:06:04 [g2]: ping Jan 23 00:11:20 <[g2]> GPSFan pong Jan 23 00:11:29 <[g2]> How goes mate ? Jan 23 00:11:47 hey, how's things... happy new year etc... Jan 23 00:12:28 <[g2]> GPSFan good thx. Jan 23 00:12:31 <[g2]> you ? Jan 23 00:13:04 <[g2]> GPSFan did you see the checkins from the last hour ? Jan 23 00:13:11 [g2]: haven't done any avila hacking since dec. just starting to spin up again, does 2.6.20-rc? work yet? I've been skiing a few times. ;>) Jan 23 00:13:24 yes, I just read the logs. Jan 23 00:13:45 <[g2]> .20-rc is unhappy for the avila/loft due to the PATA Jan 23 00:14:04 <[g2]> it's also unhappy due to some PCI issues Jan 23 00:14:15 ah, ;.( Jan 23 00:14:18 <[g2]> but .19 is quite happy Jan 23 00:14:44 <[g2]> I don't know of any real outstanding issues with .19 Jan 23 00:15:17 I got all my code ported and it runs ok. 2.6.19-0 works quite well, I'll have to get -2 running in the near future. Jan 23 00:16:40 I see that Christian got some of the crypto stuff going too. Jan 23 00:16:54 <[g2]> I haven't ever tried that but it looks very interesting Jan 23 00:17:32 indeed, is there any userland to go with it? or is that still to be? Jan 23 00:17:40 <[g2]> I think it's going to be a little while until .20 gets straightened out Jan 23 00:18:06 <[g2]> dunno about the crypto Jan 23 00:18:06 well as well as .19 works, I'm not in a hurry. Jan 23 00:18:41 <[g2]> GPSFan are you following l-a-k ? Jan 23 00:19:01 [g2]: I scan it every few days. Jan 23 00:20:15 ooh. lots of activity today.. have to catch up.. Jan 23 00:21:52 <[g2]> last couple days :) Jan 23 00:23:38 well, that shows what happens wehn I go off to work on another project... ;>) (or go skiing...) Jan 23 00:43:46 <[g2]> GPSFan enjoy the skiiing! :) Jan 23 00:48:46 [g2]: thanks, hope to go again later this week. cul off to dinner... Jan 23 00:49:06 <[g2]> GPSFan cheers **** ENDING LOGGING AT Tue Jan 23 02:59:59 2007