**** BEGIN LOGGING AT Fri Aug 30 03:00:27 2019 Aug 30 05:43:56 i wanted to add "system" new ubus method "check_image" but it not clear enogh Aug 30 05:44:06 what do you find better: "check_fw" or "check_firmware" ? Aug 30 05:46:51 check_firmware Aug 30 06:00:32 check_firmware_image Aug 30 06:01:10 actually verify_firmware_image Aug 30 06:01:23 check is too generic Aug 30 06:01:47 as is just firmware, since its unclear whether it applies to the running system or a given image file Aug 30 06:02:11 i see Aug 30 06:02:56 how does it work in practise, will you pass a path as argument to the call? Like ubus call system verify_firmware_image '{ "path": "/tmp/image.bin" }' ? Aug 30 06:03:04 yes Aug 30 06:03:10 ok Aug 30 06:05:12 i see we use "check" everywhere Aug 30 06:05:24 platform_check_image, fwtool_check_image, fwtool_check_signature Aug 30 06:05:28 $sysupgrade_image_check Aug 30 06:05:40 echo "Image check '$check' failed." Aug 30 06:05:46 so... maybe check_firmware_image ? Aug 30 06:06:00 i think it's still a clear name Aug 30 07:29:25 blogic: could you review my procd patch, please? you couldn't even see it yet, it's just a head-up Aug 30 08:20:52 rmilecki: ping Aug 30 10:07:29 blogic my procd patch too, maybe? Aug 30 11:24:16 Hi. I am currently trying to add support for a QCA9558-based device to OpenWRT. The device closely resembles the Comfast WR650AC (my device is also Comfast), so I am using that device as a starting point Aug 30 11:25:38 However, on boot I get a kernel oops in ar8327_cleanup (so most likely ar8327_leds_cleanup). If I disable CONFIG_AR8216_PHY_LEDS, then the board boots find and Ethernet works as intended Aug 30 11:25:46 Does anyone have any suggestions as for what could be wrong? Aug 30 11:42:17 kristrev: any more details i nthe oops? Aug 30 11:44:36 jow: I somehow forgot to finish the message containing the pastebin-link containing the oops. Here it is: https://pastebin.com/5V7tUUKz Aug 30 11:46:15 kristrev: so it looks as if it gets freed because the probe function didn't find suitable hardware Aug 30 11:47:12 kristrev: I'm not expert at all with this code but from reading ar8327.c I think the issue could be that priv->chip_data is simply NULL in this case Aug 30 11:48:02 kristrev: it will only get set to allocated memory in ar8327_hw_init() which likely won't get called when the probe fails Aug 30 11:48:28 Yes, agree. I suspect the problem might be the call to kfree(data->leds); Aug 30 11:48:35 even earlier Aug 30 11:48:49 struct ar8327_data *data = priv->chip_data; Aug 30 11:48:50 ... Aug 30 11:48:56 data->num_leds Aug 30 11:49:04 this already is a null pointer access Aug 30 11:49:40 I think the entire function should bail out on data == NULL Aug 30 11:51:55 if chip_data was NULL, then to me it seems that driver init will fail with ENOMEM and the switch will not work Aug 30 11:52:38 I am sprinkling some printks to see if some debug info will reveal what goes wrong Aug 30 11:52:45 your oops trace showed mdio_probe -> ar8xxx_mdiodev_probe -> ar8xxx_free -> ar8327_cleanup Aug 30 11:53:06 this looks like at least some instance of the driver got initialized, failed during probe, then freed again Aug 30 11:53:22 ah, true, I agree Aug 30 11:53:27 thanks for spotting that Aug 30 11:56:07 yw Aug 30 11:56:55 I also see that if switch was properly initialized, there should have been a message "... switch registered ..." Aug 30 11:56:59 Time to keep digging Aug 30 12:36:21 hmm, during x86/64 build I see this: warning: Cannot use CONFIG_STACK_VALIDATION=y, please install libelf-dev, libelf-devel or elfutils-libelf-devel Aug 30 12:36:40 looks to me like a new implicit host dependency silently crept in Aug 30 13:11:38 meh, that entire thing is foobar Aug 30 13:11:45 it uses host libelf, not target libelf Aug 30 13:11:53 or rather tools/ libelf Aug 30 13:12:09 when that is fixed, it fails to build because tools/libelf/ has broken headers Aug 30 13:28:11 Hi guys. I'm new in qsdk and I need help. In my router I have hostapd with driver=atheros in config file. As far as I look into code of hostapd and my atheros card I don't see an option to use WPA3 as authentication. Is there any possibility to do that or it is not yet implemented? Aug 30 13:40:31 Halfboy: wpa3 is possible with some prerequirements, AFAIK https://openwrt.org/docs/guide-user/network/wifi/basic Aug 30 13:40:43 ..has more relevant info Aug 30 13:43:04 Thanks olmari. But is it applicable for atheros driver ? Where the authentication is done in code? Aug 30 13:43:15 halfboy: very unlikely Aug 30 13:44:30 Mm.. I kinda missed that point... atheros is old as.. guano Aug 30 13:46:14 halfboy: if you use qsdk you probably need to ask qualcomm for support; qsdk has a lot of differences from openwrt Aug 30 13:47:01 Thanks! Aug 30 13:47:16 Is it better alternative than atheros? Aug 30 13:47:22 olmari: it's not old, its just the closed source wifi driver Aug 30 13:48:00 Okay today is obviously not my day then xD Aug 30 13:49:25 kristrev: sounds like you found a bug in an error path ;) Aug 30 13:50:45 KanjiMonster: I think I mistought madwifi somehow Aug 30 13:51:51 I saw in nl80211 flags like _SAE_. Is it connected with wpa3 stuff? Aug 30 13:54:18 yes Aug 30 13:54:21 SAE, OWE Aug 30 13:56:25 Hauke: Are there any plans to go ahead with https://patchwork.ozlabs.org/patch/1095260/ or even update hostapd to version 2.9? Aug 30 14:06:28 jow: kristrev: chip_data is allocated in ar8xxx_start -> ar8327_hw_init; if register_switch fails, it never calls ar8xxx_start, and then the null pointer access happens in ar8xxx_free -> ar8327_cleanup Aug 30 14:08:56 or if probe_switch fails Aug 30 14:14:05 * f00b4r0 stumbles upon https://bugs.openwrt.org/index.php?do=details&task_id=1764 after a failed FTP connection, wonders why kmod-ipt-raw isn't included by default Aug 30 14:14:50 to remind you not to use ftp? :) Aug 30 14:15:25 heh :) Aug 30 14:21:48 jow: what uses cgi-io? I'd never even heard/seen it before Aug 30 14:22:14 hmm. fw3 still doesn't populate the raw table, wth Aug 30 14:30:18 jow: pong Aug 30 14:30:27 KanjiMonster: Hehe. Turns out that I should never have invoked the code though :) Seems there is no switch on the device, on a single interface (I am probably using the wrong terms) Aug 30 14:31:23 After removing the switch-node and only enabling mdio + leving eth0 then network + device works like a charm Aug 30 14:55:37 kristrev: that crash might be worth fixing anyway Aug 30 14:55:43 karlp: luci will use it soon Aug 30 14:56:04 rmilecki: uhttpd.init does not require openssl, its just one of the two choices to generate certs on startup, the other is px5g Aug 30 14:56:18 rmilecki: or if you have other means to stage certificate files, you need neither Aug 30 14:58:23 so I should probably just select openssl-utils/px5g manually in my .config (e.g. using make menuconfig)? Aug 30 14:59:34 right Aug 30 15:00:00 ok, thanks Aug 30 15:00:16 rmilecki: or consider shipping files/etc/uhttpd.crt && files/etc/uhttpd.key Aug 30 15:01:06 cert generation on target will take a lot of time and will suffer from bad entropy anyway, so you might as well pregenerate selfsigned certs on your build host+ Aug 30 15:02:24 interesting point, i didn't think about time & entropy Aug 30 15:14:31 jow: Agree Aug 30 15:57:59 I will take a look at fixing the error when I have some free time, unless someone beats me to it :) Aug 30 16:07:19 what's the command name to put down a symlink in the filesystem in the install section of a package makefile? (is this actually the right channel to ask these questions as in asking questions related to developing with in the openwrt environment or should I rather go to #openwrt with this kind of questions?) Aug 30 16:21:37 x5a17ed: $(LN) ? Aug 30 16:22:32 x5a17ed: find ./package/ -name Makefile -exec grep -H "LN" {} \; Aug 30 16:24:03 I tried ~/git/openwrt.org/packages $ ag '$(LN)' which gave me nothing Aug 30 16:24:07 but thanks! Aug 30 16:24:49 oh wait, I see the error, I had to escape the parentheses Aug 30 16:24:57 i'm not familiar with ag, but try \$(LN) Aug 30 16:25:02 right Aug 30 16:25:13 my bad (－‸ლ) Aug 30 16:29:17 rmilecki: ack LN --type=make :) Aug 30 16:35:46 usb camera /dev/video0: fd=open(/dev/video0); unplug camera; /dev/video0 file gone immediately; can I still close(fd)? Aug 30 16:48:48 Pepe: DFS with ieee80211a is still broken in hostapd 2.8 / 2.9 Aug 30 16:48:54 I haevn't found the time to debug this Aug 30 16:55:34 @ausjke, yes, the fd stays to be valid but io on it will fail (might be mistaken though, test it out) Aug 30 17:38:49 jow: do you manage the mailing list? Aug 30 17:46:28 jow: ping Aug 30 21:20:14 any hints on how to bypass the following ipq806x build error on master? Aug 30 21:20:21 * check_data_file_clashes: Package ath10k-firmware-qca99x0-ct wants to install\ Aug 30 21:20:23 file /home/n/openwrt/build_dir/target-arm_cortex-a15+neon-vfpv4_musl_eabi/linu\ Aug 30 21:20:26 x-ipq806x_generic/target-dir-261ef631/lib/firmware/ath10k/QCA99X0/hw2.0/board-2\ Aug 30 21:20:27 .bin Aug 30 21:20:29 But that file is already provided by package * ath10k-firmware-qca99x0\ Aug 30 21:20:32 -ct-htty Aug 30 21:20:39 don't include both of them? Aug 30 21:21:13 karlp: ct-htt is set to y in .config, ct only is set to m Aug 30 21:21:24 karlp: can't unselect ct only... Aug 30 21:21:49 tired hand editing .config but that didn't seem to work either... Aug 30 21:22:16 press ? in menuyconfig and see what's holding it selected then. Aug 30 21:22:24 do I need to make clean after manual chaning .config? Aug 30 21:22:29 k, I'll take a look Aug 30 21:22:42 manually changing .config is always error prone. Aug 30 21:22:50 at least make defconfig to fill in pieces again. Aug 30 21:24:35 Selected by: MODULE_DEFAULT_ath10k-firmware-qca99x0-ct [=y] && \ │ Aug 30 21:24:38 │ TARGET_PER_DEVICE_ROOTFS [=y] && m && MODULES [=y]y Aug 30 21:25:02 shouldn't need ct only if using htt... Aug 30 21:25:46 hint about location of makefile dep for ath10k? Aug 30 21:26:05 s/location/locations/ Aug 30 21:28:02 I already tried setting ct to ct-htt in /target/linux/ipq806x/image/Makefile Aug 30 21:36:05 nmrh: I'm using this locally http://paste.debian.net/hidden/e59fc22c/ Aug 30 21:37:33 pkgadd: thx, thats kind of what mine looks like but only for r7500v2 Aug 30 21:37:47 guess I need to make clean or something Aug 30 21:38:07 silly to recompile all of it tho Aug 30 21:38:19 'shouldn't' be necessary (famous last words), but I'm using that for nbg6817 and BThub5 Aug 30 21:38:30 k, I'll try again... Aug 30 21:38:39 (and also merely build testing for r7800) Aug 30 21:44:56 i'll try but I suspect it won't work, even after make clean, make menuconfig, my .config looks the same (with the dep changes) Aug 30 21:45:19 it must be being set somewhere else but I didn't find it... Aug 30 21:46:34 i did copy a diffconfig to .config (with only ct-htt set) before menuconfig... Aug 30 21:47:02 I always start from a minimized diffconfig and expand it via make defconfig Aug 30 21:47:37 http://paste.debian.net/hidden/fb946cdf/ Aug 30 22:37:05 pkgadd: thx for the example, It looks like I have to spend some time redoing my diffconfig Aug 30 22:37:20 something changed over the past month and I missed it Aug 30 22:41:23 I trim that down to the minimum - and modularize it as much as possible (only ipq806x.init changing between different targets, replacable with ath79, lantiq, brcm47xx, etc.) Aug 30 22:42:50 yea, I wan't a minimum as well but I'm still on the learning curve on how to get it without messing up Aug 30 23:16:09 karlp: ping Aug 30 23:31:32 yeah man, whats up? Aug 30 23:32:08 karlp: https://github.com/openwrt/packages/pull/9595 Aug 30 23:33:20 sure, even better if they want to maintain it. Aug 30 23:33:37 I have it locally as well, but it's only used in another package, and that package I can't uðpdate due to some ssl breakage. Aug 30 23:33:53 so I didn't have any package that used it to verify it other than compile testing. **** ENDING LOGGING AT Sat Aug 31 02:59:57 2019