**** BEGIN LOGGING AT Mon Mar 11 02:59:57 2019 Mar 11 07:48:41 morning Mar 11 08:16:50 wigyori, o/ Mar 11 11:35:12 pepe2k: just want to say thanks for your u-boot, i'm using it on a bunch of tplink wdr3600's Mar 11 11:40:40 russell--: you're welcome, new version coming soon Mar 11 15:38:00 Are there any user-land tools for working with the "draft" qca8k driver at this point, even at the register level? Mar 11 15:40:42 jeffsf: whats draft about the driver Mar 11 15:40:50 and what do you want ot do from userland ? Mar 11 15:40:53 its a DSA driver Mar 11 15:41:02 Just noticed that it has been upstreamed, Mar 11 15:41:05 you can use the bridge tool and ip Mar 11 15:41:29 OK, will look into that in greater detail Mar 11 15:42:38 With the QCA tags, each port ends up on its own subinterface, and was looking into how to configure so that only those packets that the switch fabric didn't "understand" were forwarded to the CPU Mar 11 15:43:43 the switch in the default config will fwd all frames to the cpu port where the fdb lookup fails and/or the fdb lookup points at the cpu port Mar 11 15:44:11 as long as the cpu and ingress swdevport are in the same loopkup group Mar 11 15:44:16 OK, that was what I was hoping Mar 11 15:44:27 so there is full L2 offloading Mar 11 15:45:10 Right -- at least for now I'll be happy to get the IPQ4019-based device working with a single pathway between the switch and the CPU Mar 11 15:45:19 ah ok Mar 11 15:45:25 i have done some prior work on that Mar 11 15:45:45 the problem is that the dakota needs to have mmio based i/o to the switch and a special tag driver Mar 11 15:45:58 Maybe another day to enable the second pathway, assuming there's a "listener" in the IPQ4019 for it Mar 11 15:46:37 I saw "[RFC,4/4] net-next: dsa: qca8k: add support for multiple cpu ports" but haven't dug too deply into it Mar 11 15:46:57 ( https://patchwork.ozlabs.org/patch/710841/ ) Mar 11 15:47:00 that is not what you want Mar 11 15:47:02 https://github.com/blogic/linux-1/commit/11a8019b0656d203b04bcd2c0867dbf777643e95 Mar 11 15:47:18 ipq40xx has single cpu port support Mar 11 15:47:28 OK, thanks -- on both accounts Mar 11 15:47:30 that patch is only part of the solution Mar 11 15:48:03 the problem being that on TX the dst port needs to go into the dma ring and not the magic tag if i recall Mar 11 15:48:41 so the tag driver puts the info in the skb but the ethernet driver had to extract it again and put it into the dma if i recal correctly Mar 11 15:49:33 I saw that the stand-alone QCA8337N supports 2x RGMII or 1x RGMII / 1x SerDes, but don't have that level of detail of what's in the IPQ4019 Mar 11 15:50:10 ( either in the switch chip, or in what pathways are present in the SoC ) Mar 11 15:50:46 ... switch functionality on the SoC die ... Mar 11 18:49:16 blogic: the stmmacs are betweeen the CPU and the switch? Mar 11 20:32:58 no, stmmac is the gmac and connects to the cpu port of the witch Mar 11 20:33:03 *switch Mar 11 20:53:24 The more I look into these things, the more sorcery and witchcraft I believe is in them Mar 11 20:54:07 a bit more research and you can become a tech evangelist jeffsf ;) Mar 11 21:07:02 Google is funding technical writers: https://lwn.net/Articles/782785/ do we want to participate and does someone want to organize this? Mar 11 21:19:50 jeffsf: switchcraft ;-) ldir fetches coat Mar 11 21:20:08 LOL Mar 11 21:20:48 As long as I don't end up with my EA8300 becoming pitch-craft Mar 11 21:21:11 I think there's a gremlin in its U-Boot Mar 11 21:22:01 "You have entered the TWILIGHT ZONE" Mar 11 21:22:16 did you remember the rules? No water, no feeding after midnight and urrrrr Mar 11 21:22:46 No, no, you're going to boot_part 2, because I don't like boot_part 1 Mar 11 21:23:32 it boots part 4? Mar 11 21:24:06 It has a mind of its own Mar 11 21:24:30 "No, I don't like that environment. I'll just rewrite it for you." Mar 11 21:25:13 goes back to reading the c programming wiki book - the pointers chapter - AGAIN Mar 11 21:28:58 *((volatile uint32_t *)0) = 0 Mar 11 21:29:26 At least on a 32-bit address-space machine Mar 11 21:30:06 Good trick for getting an MPU into the boot loader again Mar 11 23:39:15 Hi, any commiters here? wouldnt you include patch in rb915g please? https://openwrt.org/toh/mikrotik/rb951g_2hnd its frustrating that nor stable nor snaphost openwrt works on these routers with gigabit ports Mar 11 23:40:39 I have rev. 2 and its unusable - packet drops Mar 11 23:41:45 user: https://openwrt.org/submitting-patches Mar 11 23:43:44 pkgadd: ok, then I will try to create one lol Mar 12 00:18:33 will there ever be a solution for this? Mar 12 00:18:36 [447169.464656] ath: phy0: DMA failed to stop in 10 ms AR_CR=0x00000024 AR_DIAG_SW=0x42100020 DMADBG_7=0x000286c0 Mar 12 00:19:28 LEDE_BOARD="ar71xx/generic" Mar 12 00:19:40 LEDE_RELEASE="OpenWrt 18.06.2 r7676-cddd7b4c77" Mar 12 00:19:59 system type : Atheros AR9132 rev 2 Mar 12 00:19:59 machine : TP-LINK TL-WR1043ND **** ENDING LOGGING AT Tue Mar 12 02:59:57 2019