**** BEGIN LOGGING AT Sat Aug 01 02:59:58 2009 Aug 01 09:13:45 <_trine> I have just asked on #openwrt but got no response has anyone any idea how to make svn trunk download it always used to work but now after deleting it I get this returned:- Aug 01 09:14:02 <_trine> svn: Server sent unexpected return value (405 Not Allowed) in response to OPTIONS request for 'https://svn.openwrt.org/openwrt/trunk Aug 01 09:17:01 <_trine> has the repository been moved? Aug 01 09:40:33 <_trine> solved Aug 01 09:47:11 <[florian]> _trine: ok, I was about to direct you to the page which mentions its move to sv:// Aug 01 09:49:16 <_trine> [florian], thanks I must admit I already knew about the change from https to svn but I had forgotten Aug 01 09:50:41 <_trine> bloody old age taking it's toll Aug 01 10:51:21 florian * r17072 /trunk/package/base-files/Makefile: [package] prevent rebuild issues on creation of lib64 with 64-bits targets Aug 01 11:09:04 florian * r17073 /trunk/package/rtc-rv5c386a/ (. Makefile src/ src/Makefile src/gpio.h src/rtc.c): [brcm-2.4/brcm47xx] add RTC RV5C386A package Aug 01 11:10:03 florian * r17074 /trunk/target/linux/brcm-2.4/profiles/WLHDD.mk: [brcm-2.4] update profile to include the rtc driver Aug 01 12:10:24 florian * r17075 /trunk/include/package-defaults.mk: [package] revert 17050 this was not the right way to fix this Aug 01 12:11:01 florian * r17076 /trunk/package/iptables/Makefile: [package] force iptables to be configured with ipv6 enabled (#5586) Aug 01 12:12:49 florian * r17077 /trunk/rules.mk: [package] do not set anything in DISABLE_IPV6 when set, let the package pick up the defaults Aug 01 12:18:21 [florian]: *ping* Aug 01 12:18:28 <[florian]> cshore: pong Aug 01 12:19:23 Hi, just wondering if you could let me know about kernel patches...how do I apply all the patches but not compile anything, so that I have pristine OpenWRT kernel for modifying and creating my own patches? Aug 01 12:20:02 make target/linux/{clean,prepare} V=99 Aug 01 12:20:13 however your kernel tree does not have to be pristine for creating your own patches Aug 01 12:20:14 <[florian]> damn it, you are too fast :p Aug 01 12:20:18 the kernel tree is unpacked using quilt Aug 01 12:20:21 oh? Aug 01 12:20:23 and you can edit the quilt patchstack Aug 01 12:20:30 and let the build system copy your stuff back when you're done Aug 01 12:21:10 er, need to learn quilt yet, all I know how to do is apply and unapply Aug 01 12:21:13 cshore: btw. are you daniel or is it just similarity in the email address? Aug 01 12:21:34 nope, that's me (Daniel aka cshore) Aug 01 12:21:37 ah, ok Aug 01 12:21:48 just sent you a follow-up on the last mail you sent me Aug 01 12:21:57 thnx Aug 01 12:24:04 if you want, we can continue the discussion here Aug 01 12:25:03 The problem I'm seeing is that things don't work if the start of the jffs2 moves (e.g. the squashfs is more than an erase block different than were it was before) Aug 01 12:25:33 did you try mtd -j ? Aug 01 12:25:59 because why use the refresh if mtd can integrate the data during flashing on-the-fly Aug 01 12:26:20 But doesn't that presume an existing jffs? Aug 01 12:26:23 no Aug 01 12:26:34 it presumes an end-of-filesystem marker in the image that you want to flash Aug 01 12:26:43 ah, that would be the problem then Aug 01 12:26:46 which is used as an indicator of where the data is supposed to go Aug 01 12:26:55 openwrt by defaults adds this marker to every single image Aug 01 12:27:01 even for squashfs Aug 01 12:27:36 and as long as the mtd util can figure out the erase block boundaries, it should find the right marker Aug 01 12:27:36 I need to eliminate that marker as part of the filesystem image because when it gets erased and replaced by the jffs metadata it causes the CRC to change and some routers recheck the CRC on every boot Aug 01 12:27:52 what platform? Aug 01 12:27:56 brcm63xx Aug 01 12:28:01 what kind of image format? Aug 01 12:28:06 imagetag Aug 01 12:28:14 ok, i had the same problem for trx Aug 01 12:28:18 and i added a crc fixup to mtd Aug 01 12:28:29 if you can implement a similar fixup for imagetag and make it platform specific, it should work Aug 01 12:28:39 it simply recalculates the crc of the image after flashing everything Aug 01 12:28:55 hmmm...what happens when the jffs inodes change? Aug 01 12:28:57 that's simpler and more reliable than the refresh stuff Aug 01 12:29:12 can't you limit the scope of the crc check? Aug 01 12:29:20 that's how we did it for trx Aug 01 12:29:41 it checks everything that is part of the rootfs partition Aug 01 12:30:00 where does it get the size from? Aug 01 12:30:05 (for flashing purposes) Aug 01 12:30:07 the imagetag Aug 01 12:30:09 ok Aug 01 12:30:12 then it should be easy Aug 01 12:30:25 make the fixup in mtd reduce the size of the rootfs partition Aug 01 12:30:34 or move it outside of the imagetag space entirely Aug 01 12:30:36 during flashing Aug 01 12:30:47 hmmm..... Aug 01 12:30:47 so that the boot loader only checks a limited search space Aug 01 12:31:04 how do we make that work for cfe flashing as well? Aug 01 12:31:37 and is there examples to go by? Aug 01 12:31:56 or web interface (ODM) Aug 01 12:31:58 that part is a bit messy Aug 01 12:32:01 right Aug 01 12:32:08 but we solved it for trx as well Aug 01 12:32:15 though not in a particularly beautiful way Aug 01 12:32:28 that's why I eliminated the DEADC0DE marker from the image in the patch I send to -devel Aug 01 12:32:28 the mtd map detects the partial jffs2 coverage of the header Aug 01 12:32:31 and fixes it up on first boot Aug 01 12:32:50 the problem is if you eliminate that marker you open up a whole can of different issues Aug 01 12:32:59 related to clean boot with squashfs+jffs2 Aug 01 12:33:03 right Aug 01 12:33:15 so the fixup is probably the most practical way to deal with this Aug 01 12:33:51 I solved all those issues, except the one of being able to wipe an a rootfs by flashing Aug 01 12:34:05 and I had an idea for that Aug 01 12:34:54 And it allowed for the possibility of preserving the rootfs across a firmware flash when we want that Aug 01 12:36:21 i don't think we want to preserve the rootfs across flashing Aug 01 12:36:23 A possibility I see to make -j work in though is to detected, instead of DEADC0DE, the squashfs end of filesystem marker Aug 01 12:37:09 Hmmm.....I think it could be useful, but I'm willing to be convince otherwise Aug 01 12:38:19 for -k for the specific case of a flash on a non-existing jffs2 (even DEADC0DE) Aug 01 12:38:34 then we still don't need mtd refresh Aug 01 12:39:48 except, here's a question; how does -j detect erase block boundaries (just check each boundary for the marker?) Aug 01 12:41:33 The other things is that basically the same code could be applied to using a usb root instead of mtd root (e.g. for a router with USB, like mine) Aug 01 12:42:42 That and the CRC issue are why I wanted to eliminated DEADC0DE as the check for a filesystem that is from a different firmware than the current squashfs Aug 01 12:45:30 -j detects the erase blocks by querying the mtd device Aug 01 12:46:21 and then checks each erase block for the marker? Aug 01 12:47:07 it looks at the offset Aug 01 12:47:49 Ok, I'm confused...if the offset and size change, how would that work (or does it base it of the *new* settings?) Aug 01 12:48:03 it flashes the image as one complete image Aug 01 12:48:17 so it doesn't care about the overlapping sub-partitions Aug 01 12:49:12 that's the reason why it doesn't need the refresh Aug 01 12:49:19 because it doesn't care about the updated split Aug 01 12:51:01 I admit I'm still confused. Does it determine the squashfs by looking at where the image ends. Or reading the imagetag? Basically how does it know where the jffs starts? Aug 01 12:51:34 Never mind Aug 01 12:51:41 I think I got it Aug 01 12:52:18 lars * r17078 /trunk/target/linux/s3c24xx/files-2.6.30/drivers/mfd/glamo/glamo-gpio.c: [s3c24xx] Add glamo gpiolib support. Aug 01 12:52:53 To determine the jffs start it examines the image for the marker, and then determines *where to put it* by querying erase block boundaries, and putting it on the first erase block after the image. Right? Aug 01 12:54:14 almost Aug 01 12:54:27 it checks the start of each erase block that it writes Aug 01 12:54:37 by querying the boundaries of the chip before starting Aug 01 12:54:49 when an erase block starts with 0xdeadc0de, it replaces that with the jffs2 data Aug 01 12:54:56 and then adds a new marker after that Aug 01 12:55:17 hmmmm....okay Aug 01 12:56:24 That eliminates the need for part of the patch I did Aug 01 12:58:10 I still think not relying on DEADC0DE to determine whether to format the jffs2 (e.g. for a CFE flash) is a good idea, because of the USB root I want to do (which will need something else anyway), and I'm partial to preserving the rootfs, if that's the choice we want to make at a particular time Aug 01 12:58:41 Bit I see how to do that with a -j type parameter and the squashfs magic Aug 01 13:00:35 i think even with usb root, the same trick that i use on x86 might work Aug 01 13:00:43 what's that? Aug 01 13:00:50 block2mtd on the block device Aug 01 13:01:05 and block2mtd contains a refresh call, which refreshes the underlying partition table Aug 01 13:01:59 Hmmm....but does that require a different target in the compile system (I want to have the choice of mtd or usb root for brcm63xx) Aug 01 13:03:08 Basically is there a clean way to implement the only difference being the use of block2mtd? Aug 01 13:04:24 i intend to use aufs for external rootfs systems at some point Aug 01 13:04:32 i.e. the system boots from flash, reads the config Aug 01 13:04:38 waits for the external system to come up Aug 01 13:04:45 and then integrates it into the overlay chain Aug 01 13:05:19 You mean using aufs instead of mini_fo? Aug 01 13:05:22 yes Aug 01 13:05:42 Because that can combine more than just once, right? Aug 01 13:06:37 yes Aug 01 13:07:05 hmmm..... Aug 01 13:09:49 Can you edit the config on the mtd if the aufs has the external fs overlay on top of it? That would be perfect if it's the case (can you have two rw layers?) Aug 01 13:10:41 with aufs you can select which fs you want the overlay stuff to be written to Aug 01 13:11:01 so the initial writable layer is the flash filesystem Aug 01 13:11:23 but as soon as it has mounted the external media, it switches Aug 01 13:12:50 can you set it up so that existing files on the underlying (flash fs) are rewritten if changed, but the overlay is where new files go? Aug 01 13:14:14 basically can you change the boot config without dropping to failsafe? Aug 01 13:18:14 yes Aug 01 13:30:44 Ok, so most of the patch I did was unnecessary Aug 01 13:35:00 hmmm.....I'm thinking about it though and the CRC is still a problem Aug 01 13:35:19 lars * r17079 /trunk/target/linux/generic-2.6/patches-2.6.30/780-fix-race-in-snd_soc_jack_add_gpios.patch: fix patch Aug 01 13:37:28 If flashed via web interface we would have to somehow rewrite the imagetag in order for the CRC to be correct because the web interface requires the rootfsoffset be the start of the image and the kernelLength+rootfsLength = length of image to flash Aug 01 13:37:59 so we can't exclude DEADC0DE and still have it written to the flash by the ODM web if Aug 01 13:38:23 if it's part of the flash image Aug 01 13:44:36 <_trine> I think I may have found something wrong with radvd can someone check it for me Aug 01 13:45:54 <_trine> radvd seems to think that the location should be /etc/radvd.conf when in fact it is at /etc/config/radvd.conf Aug 01 14:01:55 hi all! Aug 01 14:03:12 <_trine> update radvd is actualy located at /etc/config and not radvd.conf Aug 01 16:44:17 florian * r17080 /trunk/target/linux/generic-2.6/patches-2.6.30/998-openwrt_lzma_options.patch: [kernel] also use openwrt-specific lzma options in gen_initramfs_list.sh Aug 01 16:45:13 [florian] Hello! Aug 01 16:45:40 <[florian]> hi AndyIL Aug 01 16:46:31 <[florian]> intresting problem https://forum.openwrt.org/viewtopic.php?pid=92173#p92173 Aug 01 16:47:42 <[florian]> AndyIL: no it's not interesting Aug 01 16:48:06 <[florian]> AndyIL: the problem is the alias of the flash that the bootloader has setup Aug 01 16:50:51 <_trine> [florian]: did you see my note about radvd? Aug 01 16:51:01 <[florian]> _trine: no sorry Aug 01 16:51:19 <_trine> there is a small problem that's stopping it working Aug 01 16:51:31 <[florian]> _trine: radvd should be configured with /etc/config/radvd Aug 01 16:51:43 <_trine> it's looking apparently for the wrong file in the wrong place Aug 01 16:51:55 <_trine> yes that's right Aug 01 16:52:02 <[florian]> well I have never been okay with that ucification of radvd Aug 01 16:52:04 <_trine> but it's not working Aug 01 16:52:19 <[florian]> why am I not surprised :) Aug 01 16:52:22 <_trine> when I just do radvd from cli it tells me Aug 01 16:52:40 <_trine> that it's looking for /etc/radvd.conf Aug 01 16:53:10 <[florian]> how is it started from the init script ? Aug 01 16:53:12 <_trine> yes it uci that's caused it Aug 01 16:53:54 <_trine> [florian]: I'm not sure Aug 01 16:54:23 I assume the init script generates a temporary radvd.conf (/var/etc) and then passes that to radvd Aug 01 16:56:38 florian * r17081 /trunk/target/linux/brcm63xx/ (11 files in 2 dirs): [brcm63xx] add experimental support for 2.6.30.4 Aug 01 16:57:33 <_trine> yes /etc/init.d Aug 01 16:57:44 <_trine> yes /etc/init.d/radvd Aug 01 16:59:07 so yes, one could patch radvd to use /var/etc/radvd.conf per default or something similar but it still has to be generated then somehow Aug 01 16:59:45 <_trine> http://pastebin.com/m6c41f585 Aug 01 16:59:51 <_trine> that's the bit I think Aug 01 17:02:03 iirc it only thinks to use /etc/radvd.conf is the generated config is empty Aug 01 17:02:04 <_trine> without that fixed ipv6 will never work Aug 01 17:02:33 <_trine> it was working fine in r 15554 Aug 01 17:02:38 generated config is empty if there are either syntax errors in /etc/config/radvd or if you told it to use a traditional config Aug 01 17:03:21 <_trine> xMff: I'm not sure what you mean about a traditional config Aug 01 17:03:47 traditional = skip uci foo and use /etc/radvd.conf straight away Aug 01 17:04:00 <_trine> I'm using the /etc/config/radvd I have always used Aug 01 17:04:09 then check that for syntax errors Aug 01 17:04:21 <_trine> it's always been working Aug 01 17:04:52 <_trine> I have just used the bacup I made to restore to the new firmware Aug 01 17:05:04 <_trine> so it should be ok Aug 01 17:06:15 last change to the init script was r15247 Aug 01 17:06:47 are you sure that the radvd init script is enabled? Aug 01 17:07:52 <_trine> do you want me to check something Aug 01 17:08:38 I don't see the problem actually Aug 01 17:08:47 you're supposed to use the init script Aug 01 17:09:00 if you invoke it manually you have to make sure to provide a valid config file Aug 01 17:09:31 <_trine> you mean by /etc/init.d/radvd start Aug 01 17:09:53 <_trine> but then you would have to do that each time Aug 01 17:12:05 <_trine> as a result of this there are no ipv6 interfaces when you do ifconfig Aug 01 17:12:46 <_trine> anyway I'm sure it will get sorted sooner or later now we know about it Aug 01 17:12:46 therefore I asked whether you made sure that the init script is enabled Aug 01 17:12:58 sop it runs automatically on boot Aug 01 17:13:05 <_trine> xMff: where do i look Aug 01 17:13:18 /etc/init.d/radvd enabled && echo "Is enabled" Aug 01 17:13:36 if not, run /etc/init.d/radvd enable Aug 01 17:13:40 (without d) Aug 01 17:14:48 if you installed radvd with opkg then it's most likely disabled by default Aug 01 17:15:00 <_trine> I compiled it in Aug 01 17:15:24 then it could be that your issue isn't radvd but another one Aug 01 17:15:38 there seems to be a bug that ipv6 addresses arent configured on boot sometimes Aug 01 17:16:02 <_trine> I'm not proficient at debugging all I can say really is it's not working Aug 01 17:16:31 well you could reboot, then check whether /var/etc/radvd.conf is present and whether you see ipv6 addresses configured Aug 01 17:16:51 if the config is present and the ipv6 addresses are not then your issue isn't radvd Aug 01 17:18:26 <_trine> I will reboot Aug 01 17:21:00 <__trine> yes /var/etc/radvd.conf is there and so is my ipv6 address Aug 01 17:21:28 <__trine> but there are no ipv6 interfaces Aug 01 17:21:36 <__trine> in ifconfig Aug 01 17:22:19 now do /etc/init.d/network restart Aug 01 17:22:26 and recheck Aug 01 17:23:07 <__trine> ok still no ipv6 iface Aug 01 17:24:21 __trine: radvd doesn't create an ipv6 interface Aug 01 17:24:33 it just advertises a router Aug 01 17:25:03 <__trine> no but I thought it was a radvd problem but maybe it's something else Aug 01 17:25:37 also ipv6 doesn't have a separate interface, eth0 will get an ipv6 address if you cionfigure it like that (in /etc/config/network) Aug 01 17:26:02 <__trine> I know from past experience that you do get more iface's Aug 01 17:26:14 [19:21:28] <__trine> but there are no ipv6 interfaces Aug 01 17:26:24 erx. sry Aug 01 17:26:28 stupid copy&paste Aug 01 17:26:37 no, you don#t get additional interfaces Aug 01 17:26:49 <__trine> guys i can only sy for sure it isn't working Aug 01 17:27:03 if you use tunnel tools you might, but not if you configure your ipv6 statically Aug 01 17:28:59 <__trine> my ipv6 address is showing in the other ifaces so it's trying to work Aug 01 17:29:31 <__trine> just got to go far 5 minutes Aug 01 17:29:42 <__trine> be back soon Aug 01 17:36:03 <_trine> back Aug 01 21:03:07 powerpc-linux-gcc-4.1.2: error trying to exec 'cc1': execvp: No such file or directory Aug 01 21:03:12 any idea why? Aug 01 21:03:46 possibly cuz the file doesn't exit ??? Aug 01 21:04:01 groz duh, and why is that? :) Aug 01 21:04:11 aaq-: what command are you running? Aug 01 21:04:13 oh, well, now you ask a harder question Aug 01 21:04:15 lol Aug 01 21:04:20 the ffirst one was easy Aug 01 21:23:42 <[florian]> aaq-: I actually pointed him at the ImageBuilder instead of SDK, so it was not a full-blown toolchain :p Aug 01 21:29:02 nbd * r17082 /trunk/rules.mk: add -L$(TOOLCHAIN_DIR)/lib to the LIBGCC_S variable, if necessary Aug 01 21:29:07 nbd * r17083 /trunk/target/linux/ixp4xx/ (2 files in 2 dirs): ixp4xx: add support for jumbo frames to the ethernet driver Aug 01 21:29:11 nbd * r17084 /trunk/package/madwifi/patches/106-get_arch.patch: madwifi: remove the get_arch.mk build failure when the 'detected' arch doesn't match the build system's value Aug 01 21:29:16 nbd * r17085 /trunk/package/madwifi/patches/200-no_debug.patch: madwifi: disable debug by default in the source tree instead of only relying on the override (useful for external builds) Aug 01 21:29:20 nbd * r17086 /trunk/package/madwifi/patches/370-wdsvap.patch: madwifi: fix a race condition in the ap2ap wds setup Aug 01 21:29:48 florian * r17087 /trunk/package/ebtables/Makefile: [package] update ebtables to 2.0.9-1, fixes #5001 Aug 01 21:36:23 florian * r17088 /trunk/target/linux/au1000/patches-2.6.30/007-gpio_request_button.patch: [au1000] prevent error messages on the requesting of the GPIO buttons Aug 01 23:16:08 florian * r17089 /trunk/package/iptables/Makefile: [package] install libiptc/libipulog from iptables (#5549) Aug 01 23:18:24 florian * r17090 /packages/net/iptables-snmp/ (Makefile patches/ patches/001-use_libiptc.patch): [package] unbreak iptables-snmp compilation (#5549) Aug 01 23:40:42 florian * r17091 /packages/libs/net-snmp/patches/160-no_ldconfig.patch: [package] do not run ldconfig at the end of net-snmp compilation **** ENDING LOGGING AT Sun Aug 02 02:59:57 2009