**** BEGIN LOGGING AT Thu Nov 29 02:59:56 2007 Nov 29 08:09:33 juhosg * r9627 /packages/libs/libaio/ (5 files in 2 dirs): [packages] libs/libaio: add Linux kernel AIO interface access library Nov 29 08:27:10 juhosg * r9628 /packages/utils/gadgetfs-test/ (4 files in 2 dirs): [packages] utils/gadgetfs-test: add test program for gadgetfs from linux-usb.org Nov 29 08:40:27 ejka; is the extra information recently added to https://dev.openwrt.org/ticket/2569 of any use in debugging this one? Nov 29 08:45:46 yes Nov 29 08:50:12 oliver@opencloud.com has been much more successful than me at debugging this further, but as aways... if you need any testing etc, drop me as message Nov 29 08:52:01 juhosg * r9629 /trunk/ (include/kernel-version.mk target/linux/adm5120/Makefile): [adm5120] switch to 2.6.23.9 Nov 29 11:42:56 juhosg * r9630 /packages/utils/gadgetfs-test/Makefile: [packages] utils/gadgetfs-test: fix PKG_BUILD_DIR Nov 29 18:28:06 nbd * r9631 /packages/utils/gzip/ (. Makefile): Nov 29 18:28:06 Add gzip package. Nov 29 18:28:06 Quicker than busybox's implementation. Nov 29 18:28:06 Signed-off-by: Fabian Deutsch Nov 29 19:06:29 I looked at the ppp package, specifically /lib/network.ppp.sh, and noticed it's prepared to support a "classic" PPP setup which is what I'm looking for Nov 29 19:06:53 "classic" meaning not pppoe or pptp etc. but plain ppp over a tty Nov 29 19:07:18 however it apparantly seems to yet lack a means to provide the tty to establish the connection over and I was wondering what would be the right way to pass that parameter Nov 29 19:07:39 right now $device is passed Nov 29 19:08:13 but I can't seem to use that as any UCI-configured value is overwritten in scan_interfaces() Nov 29 19:09:03 naturally I want to hack it in a way suitable to make it into the distro, that's why I'm asking for a short comment Nov 29 19:12:40 let me check Nov 29 19:13:31 assuming that "device" has a fixed purpose, as in your docs, and we don't want to touch network.sh, I thought about adding "option tty" Nov 29 19:15:18 would this help? http://nbd.name/device.diff Nov 29 19:17:23 either way, adding option tty seems to be the right way Nov 29 19:18:03 albeit it would serve a similar purpose as "device" I guess Nov 29 19:18:48 with this patch it should be possible to use option device as well Nov 29 19:19:17 if there is no way to prescribe pppd that it should create a "ppp1", yes Nov 29 19:19:21 the advantage to option tty is that it makes it explicit, that a tty is requested and the code isn't used by a wrapper Nov 29 19:20:38 and if I see it right the only way to force pppd to create ppp1 is by specifying "unit 1", so no sense in specifying an "ifname" Nov 29 19:21:05 the ifname should be set by the scanning code Nov 29 19:21:11 because other scripts expect it that way Nov 29 19:21:17 but right Nov 29 19:21:31 the user doesn't have to specify any ifname in the tty case Nov 29 19:21:50 you mean we don't support "option ifname" anyway safe for the wan port? Nov 29 19:21:53 and the ppp script could override the device variable with the contents of tty Nov 29 19:22:02 cause one could also parse a probably present "ifname" var to auto-set unit Nov 29 19:22:11 ah Nov 29 19:22:12 then my patch would probably be unnecessary and it would be safe Nov 29 19:24:19 sounds reasonable. other question is whether we want to blowup option space with another option where "device" could just be used as well Nov 29 19:24:46 let me check Nov 29 19:25:11 in the same line as the option space argument, I'd suggest to rather let users specify an "ifname" for ppp devs as well and have the script extract the unit number from that Nov 29 19:25:18 actually there's a way to make it simpler, yes Nov 29 19:25:29 however then your patch wouldn't work out :) Nov 29 19:25:33 the pppoe.sh, pppoa.sh and pptp.sh wrappers could set their own device option Nov 29 19:25:40 which is used in the same place where a real tty option is used Nov 29 19:25:53 e.g. for pppoe the device option is nic-ethX Nov 29 19:26:03 I thought they do already Nov 29 19:26:09 not yet Nov 29 19:26:15 no, wait Nov 29 19:26:19 that doesn't really fit Nov 29 19:26:29 since for pptp, device is supposed to be the actual ethernet interface Nov 29 19:26:43 = unmodified $device Nov 29 19:26:45 the firewall script expects it that way, because it wants to lock down both the ip interface and the ppp interface Nov 29 19:26:45 right? Nov 29 19:27:00 so let's do it like this: Nov 29 19:27:12 for the wrappers we leave option device as it is Nov 29 19:27:37 for ppp, we do an explicit check for proto = ppp and add the device option to the command line in that case Nov 29 19:27:49 then it's transparent for the user and the device option contains the correct value in all cases Nov 29 19:27:56 just one catch: Nov 29 19:28:16 actually we do already pass $device Nov 29 19:28:17 but Nov 29 19:28:29 $device is set to "ppp1" by means of scan_interfaces() in config.sh Nov 29 19:28:55 if it's doing that, then for some reason the scan is run twice Nov 29 19:29:01 normally the order should be like this: Nov 29 19:29:08 grab ifname, store it in device Nov 29 19:29:20 redirect ifname to match the interface that the actual ip traffic happens on Nov 29 19:30:10 I think that order _is_ maintained: however "ifname" is "ppp1" in my case, cause I thought that's the way to say what the device should be called like Nov 29 19:30:30 but even if I leave it off Nov 29 19:30:59 it seems to be set to ppp1 *confused* Nov 29 19:31:33 or do it like this, make it exactly like the other interface types: put the tty in the ifname variable Nov 29 19:31:44 then it gets copied to device by that means Nov 29 19:32:20 ouch, too easy :) Nov 29 19:32:38 safe for the fact that it's not really an "ifname", that's the right way to configure a ppp link then, I guess Nov 29 19:36:58 Fatus: is the link speed for the ppp link required on the pppd command line? Nov 29 19:37:03 one more question, what should such an iface be called like, that doesn't exist at "ifup" time, to avoid the "ifconfig: SIOCGIFFLAGS: No such device" message? Nov 29 19:37:26 nbd: no I think it must be set with stty Nov 29 19:37:52 so it's just the device name with nothing after it on the command line? Nov 29 19:38:13 if I remember correctly: yes. but I'll test. Nov 29 19:38:20 i'm making a patch at the moment Nov 29 19:40:04 ok cool Nov 29 19:40:20 how about this: http://nbd.name/ppp.diff Nov 29 19:43:46 I think you're going to wrong direction, cause you're appending $device another time. it is already appended through setup_interface_ppp() Nov 29 19:44:03 plus, device would not contain the right tty, except if specified through "option ifname" Nov 29 19:44:12 s/right tty/right device/ Nov 29 19:44:36 oh, right Nov 29 19:44:40 i missed that Nov 29 19:44:52 * nbd is too tired Nov 29 19:44:56 basically there's just two issues: Nov 29 19:45:06 ok, then just the -e instead of -f Nov 29 19:45:35 or make that one: do we want to pass the tty in "option ifname", even if it's no interface. semantics vs. pragmatism I guess Nov 29 19:47:54 what's the "-e instead of -f" part for? Nov 29 19:48:14 -e means exists, -f means exists and is a regular file Nov 29 19:48:28 might fix some bogus ifconfig commands Nov 29 19:48:29 not sure Nov 29 19:48:31 to avoid the "no such device" errmsg? Nov 29 19:48:35 nope doesn't :) Nov 29 19:48:40 ok Nov 29 19:48:42 I'll look into it Nov 29 19:49:14 just a comment on "option ifname" plz, stay with it or invent something semantically more correct? Nov 29 19:53:06 let's stay with it first Nov 29 19:58:33 nbd: http://www.nordkommission.de/ppp.diff this fixes the SIOCGIFFLAGS Nov 29 20:06:10 should be safe to apply Nov 29 20:12:45 you said the firewall relies on $device Nov 29 20:13:07 probably it's needs an exception for pppX Nov 29 20:13:09 let's see Nov 29 20:42:30 re Nov 29 20:58:16 Fatus: http://nbd.name/ppp.diff - slightly different fix Nov 29 21:01:24 the -e was not necessary in my tests so far, btw Nov 29 21:02:12 nbd: In a package/*/Makefile, do the CONFIG_* entries refer to openwrt config entries, or to kernel config entries? (for a kernel module package) Nov 29 21:03:04 some packages include the kernel config as well Nov 29 21:03:24 so it can refer to both in that case Nov 29 21:03:32 Looks like the b43 one doesn't. Can you point me to one that does, so I can see how to do it? Nov 29 21:05:53 for what? Nov 29 21:06:46 The b43 Makefile has $(if $(CONFIG_LEDS_TRIGGERS),CONFIG_B43_LEDS=y) Nov 29 21:07:33 ... but CONFIG_LEDS_TRIGGERS isn't defined, because it's a kernel config item, not an openwrt item Nov 29 21:07:57 are you sure it's not defined? Nov 29 21:08:15 i think i remember putting an include statement for the kernel config in the kernel include somewhere Nov 29 21:09:18 Well, it doesn't seem to be. It's in my build_dir/linux-brcm47xx/linux.../.config, but the source file that depends on CONFIG_B43_LEDS is not getting built Nov 29 21:10:23 ok Nov 29 21:10:38 * nbd checks Nov 29 21:10:45 Is it perhaps being defined at a later stage? I haven't looked yet Nov 29 21:12:03 Fatus: so does the new patch work? Nov 29 21:12:27 It's needed in the processing of the Makefile, as you can see in b43/Makefile Nov 29 21:14:36 Should it be something like $(if ... $(filter ... $(KCONFIG) ... Nov 29 21:16:03 nbd: yes Nov 29 21:16:22 noz: does http://nbd.name/krnl.diff help? Nov 29 21:16:27 Fatus: good. will commit Nov 29 21:16:43 didn't test whether it configs wan however Nov 29 21:19:32 Bugger, "Some customers nationally may currently be experiencing difficulties accessing certain websites that terminate in Germany" Nov 29 21:19:42 nbd: is that you? Nov 29 21:20:10 (i.e. are you in Germany?) Nov 29 21:20:53 i'm not in germany right now, but i live there Nov 29 21:21:17 What I meant was: is nbd.name in Germany? Nov 29 21:21:20 yep Nov 29 21:21:43 * nbd copies the patch to a different place Nov 29 21:21:53 http://openwrt.pastebin.ca/802014 Nov 29 21:21:54 Ta Nov 29 21:22:32 That loads.... I'll have a look Nov 29 21:47:14 n8 Nov 29 21:51:15 nbd: No, that patch doesn't work... Still the same behaviour. Nov 29 22:02:48 ...but I don't see why not - it looked valid. Nov 29 22:04:47 unless it's because the include was in the KernelPackage define, and I actually need it defined in the wider context of the whole Makefile. Nov 29 22:05:29 no, there are no namespaces like that Nov 29 22:08:59 it's probably evaluated too late Nov 29 22:15:23 I'm trying it a different way, using $(filter and $(KCONFIG) Nov 29 23:09:22 nbd: I've confirmed that in the "body" of the makefile (outside define blocks), $(KCONFIG) is empty Nov 29 23:20:17 noz: try http://openwrt.pastebin.ca/802147 Nov 29 23:52:37 nbd: Yep - that'll do it (by analogy at least - I put it directly in package/b43/Makfile and it worked) Nov 29 23:52:46 I'll try your patch now. Nov 30 01:18:02 nbd: I bought another Asus wl-500gP and convinced a friend to install an Atheros mini-pci card. Now he wonders what firmware to load? Is there a stock kamikaze .trx that will work? **** ENDING LOGGING AT Fri Nov 30 02:59:57 2007