On 14-3-2017 14:21, Mark Asselstine wrote: > On Tuesday, March 14, 2017 1:41:02 PM EDT Arend Van Spriel wrote: >> On 14-3-2017 12:28, Johannes Berg wrote: >>>> It never came up with any projects so far. > > Myself and one of my colleagues had thought that this might be the case. On > the bright side this matches my inability to find any discussions on the > matter and that there is the possibility to get this functionality added. > >>>> I doubt that the patch >>>> below is sufficient. I suspect something more is needed. Using git >>>> blame I ended up finding these commits: >>>> >>>> a272a72 mac80211: allow using network namespaces >>> >>> This is needed in brcm drivers. > > I will have a closer look, thanks for the pointer to this patch. > >>> >>>> 463d018 cfg80211: make aware of net namespaces >>> >>> This has no impact on brcm drivers :) >>> >>>> 5061b0c mac80211: cooperate more with network namespaces >>> >>> This shouldn't be needed, you're not referring to init_net in brcm >>> drivers. >>> >>>> I think what is required from brcmfmac is to set netns for each >>>> netdev that we create to the same netns as the wiphy instance using >>>> wiphy_net(). >>> >>> Yes, like the mac80211 patch above. >>> >>>> Not sure if there is more to consider, but hopefully Johannes can >>>> comment on this although the mentioned commits have been around for a >>>> while. >>> >>> I don't think there's anything else. >>> >>>>> wiphy->flags |= WIPHY_FLAG_PS_ON_BY_DEFAULT | >>>>> >>>>> WIPHY_FLAG_OFFCHAN_TX | >>>>> >>>>> - WIPHY_FLAG_HAS_REMAIN_ON_CHANNEL; >>>>> + WIPHY_FLAG_HAS_REMAIN_ON_CHANNEL | >>>>> + WIPHY_FLAG_NETNS_OK; >>> >>> This is not sufficient, you still have to set the netns for newly >>> created netdevs. > > Ah, that is something I had not tested out. I moved an existing phy and its > associated vif and things worked as expected and also destroying the NS had > things move back as expected. I did not however create any new vifs > >> >> Thanks for confirming my suspicion. >> >> Regards, >> Arend > > Thanks for the quick pointers, not being all that familiar with the wireless > code I appreciate the discussion. I will have a closer look and see if I can > get a patch out for review. Hi Mark, I looked at it and preparing a patch. Will send it out shortly to give it a try. Regards, Arend