**** BEGIN LOGGING AT Tue Jul 16 02:59:58 2013 Jul 16 06:32:11 [florian]: pong Jul 16 06:44:52 <[florian]> ukleinek: hey Jul 16 06:45:04 <[florian]> ukleinek: could you test swconfig so far? Jul 16 07:06:24 [florian]: didn't touch anything yet, only looking yet. Jul 16 07:19:19 nbd r37365 branches/attitude_adjustment/package/mac80211/patches/615-rt2x00-fix_20mhz_clk.patch * AA: mac80211: delete the rt2x00 clk patch again Jul 16 07:22:17 * ukleinek takes his old wrt54 and looks at the capabilities of its switch. Jul 16 07:28:28 hmm, no swconfig installed Jul 16 07:58:27 luka r37366 trunk/tools/mkimage/Makefile * [tools] mkimage: fix md5sum after r37361 Jul 16 08:27:00 <[florian]> ukleinek: if you flash a recent trunk you might be able to test swconfig with your wrt54g Jul 16 08:27:19 <[florian]> ukleinek: unless it has an admtek switch, we do not have a swconfig driver for it yet Jul 16 08:30:12 [florian]: yeah, switched (haha) to looking at my netgear wndr3800. That is in production though, so I cannot play freely. Jul 16 08:30:42 [florian]: the wrt54 is too annoying, it reboots when I do $(opkg list | grep swconfig) Jul 16 08:30:56 <[florian]> it probably goes low on RAM Jul 16 08:31:03 * ukleinek nods Jul 16 08:31:05 <[florian]> Linux is no longer friendly with 16MB devices :) Jul 16 08:31:52 <[florian]> altough that does not sound like something that should happen Jul 16 08:32:22 well, if init ooms it can happen, doesn't it? Jul 16 08:33:00 or watchdogd Jul 16 08:33:05 <[florian]> yes Jul 16 08:34:14 the wndr3800 has a rtl8366s switch. The EEPROM Serial Management Interface Jul 16 08:34:21 looks suspiciously like i2c Jul 16 08:34:27 <[florian]> yes Jul 16 08:34:42 <[florian]> old admtek 6996 switches have a similar functionning Jul 16 08:35:00 <[florian]> you basically program their eeprom with your settings over some serial interface Jul 16 08:35:05 <[florian]> and then just reboot the switch Jul 16 08:35:12 <[florian]> so it fetches its updated settings Jul 16 08:36:14 build #336 of at91 is complete: Failure [failed shell_14] Build details are at http://buildbot.openwrt.org:8010/builders/at91/builds/336 Jul 16 08:37:17 people have strange ideas Jul 16 08:38:27 <[florian]> I suppose that back then internal SRAM was more expensive than some internal eeprom Jul 16 08:38:33 <[florian]> not sure how though Jul 16 09:08:46 [florian]: how does the switch framework abstraction work? Is it simply key/value based? Jul 16 09:09:00 that's my impression playing around with swconfig Jul 16 09:09:26 <[florian]> ukleinek: yes, it uses netlink to convey information between kernel and user-space Jul 16 09:09:35 hmm, no, there are integer attributes, too. Jul 16 09:10:00 <[florian]> there are netlink commands too Jul 16 09:11:13 the rtl8366 isn't even connected to an ethernet device Jul 16 09:13:39 the list I get running $(swconfig dev switch0 help) corresponds to struct switch_attrlist attr_global, attr_port, attr_vlan in struct switch_dev_ops, right? Jul 16 09:17:32 <[florian]> right Jul 16 09:40:14 nbd r37367 trunk/Config.in * Config.in: fix depends syntax Jul 16 09:45:17 nbd r37368 trunk/tools/mkimage/patches/070-avoid_ENOMEDIUM.patch * tools/mkimage: avoid a linux specific return code Jul 16 09:46:27 hmm, the wndr3800 uses a dedicated device for wan. I wonder about the upsides. Jul 16 09:46:57 <[florian]> it is usually better to have two separated ethernet MACs, one for WAN and one for LAN Jul 16 09:47:08 performance wise? Jul 16 09:47:11 <[florian]> because your bandwidth is dedicated for both interfaces Jul 16 09:47:16 <[florian]> yes, performance wise Jul 16 09:48:21 but it's less flexible. With a switch covering both wan and lan I could put a port into wan. Jul 16 09:49:52 <[florian]> well you can still more or less do that by using bridges and VLANs, but yes I agree Jul 16 09:50:11 <[florian]> KanjiMonster has a device where the wan and lan ports go by default to two separate MACs, but via the same same switch Jul 16 09:50:21 <[florian]> such that it allows for the kind of port remapping you would want to do Jul 16 09:55:19 *that* would be nice. Jul 16 09:55:40 <[florian]> unfortunately it requires proper hardware suppor Jul 16 09:55:42 <[florian]> +t Jul 16 09:58:09 ..ooOO(a patch cable between wan and one switch port would work, but reduce the number of usable ports obviously) Jul 16 09:59:58 * ukleinek fails to get the semantic of the .set_vlan_ports and .set_port_pvid callbacks Jul 16 10:00:42 * ukleinek goes back to the wiki page describing switch config Jul 16 10:06:18 ukleinek: usually you don't have to worry about pvid; it will be automatically set correctly Jul 16 10:11:11 KanjiMonster: what is the machine [florian] mentioned you have where there are two cpu macs connected to a switch? Jul 16 10:12:10 ukleinek: it's a kirkwood based router (D-Link DIR-665); linksys also has some using the same setup (E4200v2, EA4500, maybe EA3500) Jul 16 10:13:19 I think there are also some higher end octeon devices using this kind of setup for bonding Jul 16 10:13:32 none of these is currently supported in openwrt Jul 16 10:15:31 :-( Jul 16 10:16:51 and there's no driver for the switch on the kirkwood devices, so currently couldn't conifigure it anyway Jul 16 10:19:52 so a pvid is just the default vid that is used for untagged packets Jul 16 10:29:48 ukleinek: right; usually it means the vid assigned to incoming untagged packets. (which vid's tags are stripped on exit are usually defined in a separate table) Jul 16 10:37:26 jow r37369 trunk/scripts/metadata.pl Jul 16 10:37:26 metadata.pl: use alphabetical order for menuconfig categories, also fixes random ordering of symbols in .config Jul 16 10:38:32 KanjiMonster: Is the table which tags are stripped also configurable in uci? Jul 16 10:39:19 ukleinek: its the ports of the vlans; if a port number has a "t" suffix, the tag is preserved, else it is stripped Jul 16 10:43:36 KanjiMonster: ah, so if I add "0t" to vlan1's port list in http://wiki.openwrt.org/doc/uci/network/switch#the.config.sections vlan0 packets are received untagged and vlan1 packets tagged. Jul 16 10:46:58 ukleinek: internally they are usually always tagged, so you would say that vlan1's tag will stay on exit on port 0, and vlan0's will be stripped (semantics ;). the switch driver will automatically set port 0's pvid to vlan0's vid, so that incoming untagged packets are assigned to vlan0 Jul 16 10:48:07 * ukleinek nods Jul 16 10:48:25 ukleinek: fyi, on my rspro when I do WAN and LAN on eth0 with VLAN tags, it tops around 25-30Mbps forwarding Jul 16 10:48:30 which is really bad :) Jul 16 10:48:50 my Internet uplink is 120Mbps down and will be increased to 150Mbps in a few months Jul 16 10:49:23 so I'm back to using eth0 dedicated for wan Jul 16 10:50:08 cyrusff: seems that ip6hint isn't working on 1 of my 3 internal zones Jul 16 10:50:47 how would I debug that ? Jul 16 10:51:31 stintel: and with a dedicated wan mac you get 120Mbps? Jul 16 10:51:40 ukleinek: yup Jul 16 10:51:59 I didn't expect it to be that bad, though Jul 16 10:52:22 maybe something in the driver is busted, but that's beyond my knowledge :P Jul 16 11:12:08 build #327 of brcm47xx is complete: Failure [failed shell_14] Build details are at http://buildbot.openwrt.org:8010/builders/brcm47xx/builds/327 Jul 16 11:13:16 Stintel: can you be more specific please? How did you configure it and what was the result and what did you expect?the Jul 16 11:14:09 cyrusff: hold on a sec, I'll query you a link with /etc/config/network Jul 16 11:16:43 And also did you see anything in the log? Usually the log conrains warnings if a hint cannot be applied Jul 16 11:17:35 what should I grep for ? Jul 16 11:18:13 Netifd Jul 16 11:20:19 Dont see anything wrong in the config Jul 16 11:21:18 Jul 12 02:31:17 Fri Jul 12 02:31:17 2013 daemon.err netifd: Failed to assign requested subprefix of size 64 for mgmt, trying other Jul 16 11:21:30 had to go look it up on my remote syslog server Jul 16 11:23:23 getting a /56 PD from my ISP .. so anything 00-FF should work for hint right ? Jul 16 11:23:33 and what is weird, for the other two it works fine Jul 16 11:23:37 (50 and 51) Jul 16 11:27:10 Yeah Jul 16 11:28:35 Can you do an ifstatus on the pd-interface? Jul 16 11:37:03 Ah there seems to be a > instead of a >= causing two immediately following hints to failo Jul 16 11:37:14 Will fix it later Jul 16 11:39:37 ok. still need the ifstatus ? Jul 16 11:40:09 No Jul 16 11:40:10 Thx Jul 16 12:03:08 yay i managed to have the same bug in two different programs... :/ Jul 16 12:05:13 what does "MC" stand for? Jul 16 12:06:01 <[florian]> ukleinek: in what context? Jul 16 12:06:16 <[florian]> I would assume it means multicast Jul 16 12:08:52 cyrus r37370 trunk/package/network/ipv6/6relayd/Makefile * 6relayd: Fix DHCPv6-PD with continuous hints Jul 16 12:08:57 cyrus r37371 trunk/package/network/config/netifd/Makefile * netifd: Fix IPv6-prefix assignment with continuous hints Jul 16 12:10:24 cyrus r37372 branches/attitude_adjustment/package/6relayd/Makefile * AA: 6relayd: Fix DHCPv6-PD with continuous hints Jul 16 12:10:29 cyrus r37373 branches/attitude_adjustment/package/netifd/Makefile * AA: netifd: Fix IPv6-prefix assignment with continuous hints Jul 16 12:11:27 stintel: ^ Jul 16 12:15:22 cyrusff: cheers ;) Jul 16 12:23:44 [florian]: drivers/net/phy/rtl8366_smi.c has "/* Try to find an existing MC entry for this VID */" Jul 16 12:25:22 <[florian]> that's multicast yes Jul 16 12:27:07 [florian]: IP multicast? Jul 16 12:27:24 <[florian]> ethernet multicast Jul 16 12:38:58 * ukleinek fails to understand that, also he doesn't find the description of RTL8366S_VLAN_MC_BASE in the Datasheet. Jul 16 12:41:08 could also mean "member config" Jul 16 13:27:09 does someone has a manual for the RTL8366S switch? The data sheet doesn't explain the registers available. Jul 16 14:06:45 jow r37374 trunk/package/network/config/firewall/Makefile * firewall: update to git head Jul 16 14:48:08 build #290 of rb532 is complete: Failure [failed shell_12] Build details are at http://buildbot.openwrt.org:8010/builders/rb532/builds/290 Jul 16 14:52:13 build #290 of ppc44x is complete: Failure [failed shell_12] Build details are at http://buildbot.openwrt.org:8010/builders/ppc44x/builds/290 Jul 16 15:16:48 [florian]: I'm looking at "[RFC PATCH 0/3] Add DT support for fixed PHYs", wondering if that isn't enough for my purpose. Jul 16 15:18:06 <[florian]> ukleinek: oh ok Jul 16 15:20:19 [florian]: it seems our customer isn't ordering, still I wonder if I can access the mdiobus (e.g. using miitool) when a fixed link is in use. Jul 16 15:21:46 <[florian]> I think the emulation works, but it won't give you the real values of the PHY Jul 16 15:56:44 build #322 of orion is complete: Failure [failed compile_8] Build details are at http://buildbot.openwrt.org:8010/builders/orion/builds/322 Jul 16 15:59:42 build #274 of sibyte is complete: Failure [failed shell_12] Build details are at http://buildbot.openwrt.org:8010/builders/sibyte/builds/274 Jul 16 16:13:37 build #320 of atheros is complete: Success [build successful] Build details are at http://buildbot.openwrt.org:8010/builders/atheros/builds/320 Jul 16 16:15:48 build #276 of xburst is complete: Failure [failed shell_12] Build details are at http://buildbot.openwrt.org:8010/builders/xburst/builds/276 Jul 16 18:55:36 build #310 of ramips is complete: Success [build successful] Build details are at http://buildbot.openwrt.org:8010/builders/ramips/builds/310 Jul 16 19:15:47 build #311 of ramips is complete: Failure [failed shell_14] Build details are at http://buildbot.openwrt.org:8010/builders/ramips/builds/311 Jul 16 19:57:53 build #247 of iop32x is complete: Failure [failed shell_12] Build details are at http://buildbot.openwrt.org:8010/builders/iop32x/builds/247 Jul 16 22:09:50 build #284 of avr32 is complete: Failure [failed shell_14] Build details are at http://buildbot.openwrt.org:8010/builders/avr32/builds/284 Jul 16 22:15:13 tripolar r37375 packages/net/git/Makefile * [packages/git] update to 1.8.3.3 **** BEGIN LOGGING AT Tue Jul 16 22:42:58 2013 Jul 17 00:13:58 build #309 of ppc40x is complete: Success [build successful] Build details are at http://buildbot.openwrt.org:8010/builders/ppc40x/builds/309 Jul 17 02:59:39 build #270 of au1000 is complete: Success [build successful] Build details are at http://buildbot.openwrt.org:8010/builders/au1000/builds/270 **** ENDING LOGGING AT Wed Jul 17 02:59:59 2013