From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S964823AbcFQUpG (ORCPT ); Fri, 17 Jun 2016 16:45:06 -0400 Received: from mail-oi0-f65.google.com ([209.85.218.65]:36549 "EHLO mail-oi0-f65.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754901AbcFQUpD convert rfc822-to-8bit (ORCPT ); Fri, 17 Jun 2016 16:45:03 -0400 MIME-Version: 1.0 In-Reply-To: <576448C0.4080102@broadcom.com> References: <1464791845-23944-1-git-send-email-zajec5@gmail.com> <574F30D8.5000300@broadcom.com> <576448C0.4080102@broadcom.com> From: =?UTF-8?B?UmFmYcWCIE1pxYJlY2tp?= Date: Fri, 17 Jun 2016 22:44:55 +0200 Message-ID: Subject: Re: [PATCH RFC] brcmfmac: support deleting MBSS AP interfaces To: Arend van Spriel Cc: "linux-wireless@vger.kernel.org" , Brett Rudley , Arend van Spriel , "Franky (Zhenhui) Lin" , Hante Meuleman , Kalle Valo , Pieter-Paul Giesberts , Franky Lin , "open list:BROADCOM BRCM80211 IEEE802.11n WIRELESS DRIVER" , "open list:NETWORKING DRIVERS" , open list Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8BIT Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On 17 June 2016 at 21:00, Arend van Spriel wrote: > On 17-06-16 14:30, Rafał Miłecki wrote: >> On 1 June 2016 at 21:00, Arend van Spriel wrote: >>> On 01-06-16 16:36, Rafał Miłecki wrote: >>>> We already support adding extra (AP) interfaces so it also makes an >>>> obvious sense to allow deleting them. >>>> >>>> Adding a new interface is implemented by sending request to firmware for >>>> creating a new BSS and waiting for a proper event. Ideally deleting >>>> interface should be handled in a similar way. There should be a request >>>> to firmware for deleting BSS and firmware should respond with an event. >>>> >>>> Unfortunately it doesn't seem to work with recent firmwares. They never >>>> seem to delete BSS and never send BRCMF_E_IF_DEL. As a workaround this >>>> patch deletes Linux interface while keeping a track of BSSes present in >>>> a firmware. If there is request for adding a new interface this code is >>>> capable of reusing existing BSS-es. >>> >>> It is not so much an issue of recent firmware. Actually, on recent >>> firmware 7.x.y.z and higher there are other command to create *and* >>> delete additional interfaces. On the other hand we aim to support a >>> large number of devices going back to bcm4329 so we have to come up with >>> a scheme to use the new commands or fallback to old api. Let's hope we >>> can reuse much of this effort you put in. >> >> You gave me a complex puzzle there :D It took me a while to find out >> what API you meant. > > Actually, the puzzle was supposed to be for me, but I like your > curiosity and persistence in digging up the (partial) info. > >> Finally I found an interesting wlioctl.h in SDK 9.10.178.27 that gave >> me some clue. I got this SDK from ASUS RT-AC1200G+ open souce tarball. >> There are 2 interesting structs: >> >> typedef struct wl_interface_create { >> uint16 ver; /* version of this struct */ >> uint32 flags; /* flags that defines the operation */ >> struct ether_addr mac_addr; /* Optional Mac address */ >> uint32 wlc_index; /* Optional wlc index */ >> } wl_interface_create_t; >> >> typedef struct wl_interface_info { >> uint16 ver; /* version of this struct */ >> struct ether_addr mac_addr; /* MAC address of the interface */ >> char ifname[BCM_MSG_IFNAME_MAX]; /* name of interface */ >> uint8 bsscfgidx; /* source bsscfg index */ >> } wl_interface_info_t; >> >> I couldn't find any corresponding WLC_* in wlioctl_defs.h, so I guess >> I should use WLC_SET_VAR (or WLC_SET_VAR as you prefer) with some > > (huh)? anyway the api indeed uses what we call an iovar, ie. > string-based ioctl. That's what I meant, sorry for wrong naming :) >> string. Any tip what would it be? Something like >> "wl_interface_create"? Can you reveal such a small secret? > > It is "interface_create" and "interface_remove". It (almost) works, thanks! I just hit some bug in brcmfmac in handling events. It can be exposed by deleting 2 interfaces quickly, one by one, it seems. I'll debug this. -- Rafał