User Tools

Site Tools


en:developers:summits:duesseldorf-2014

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

en:developers:summits:duesseldorf-2014 [2015/01/26 09:49] (current)
Line 1: Line 1:
  
 +Proposal for Wireless Networking Micro-conference at Linux Plumbers Conference 2014 in Düsseldorf,​ Germany... ​
 +
 +[[http://​wiki.linuxplumbersconf.org/​2014:​network_wireless|http://​wiki.linuxplumbersconf.org/​2014:​network_wireless]] ​
 +
 +etherpad notes (copied from [[https://​etherpad.fr/​p/​LPC2014_Wireless|https://​etherpad.fr/​p/​LPC2014_Wireless]]) ​
 +
 +
 +<​code>​Wireless Networking Microconference Notes
 +
 +Welcome to Linux Plumbers Conference 2014.
 +
 +The structure will be short introductions to an issue or topic followed ​ by a discussion with the audience. A limit of 3 slides per  presentation is enforced to ensure focus and allocate enough time for  discussions. ​
 +
 +Please use this etherpad to take notes. Microconf leaders will be giving a summary of their microconference during the Friday afternoon closing session.
 +Please remember there is no video this year, so your notes are the only record of your microconference.
 +
 +Schedule
 +
 +1:30 -- What's missing for Miracast? -- David Herrmann
 +miracast = wireless display, essentially
 +demo - using laptop as sink (projector),​ phone as source
 +software built using wpa_s
 +few ms delay (demo with game)
 +low framerate (30fps)
 +his software also implements the source side
 +his implementation shows up the data as a separate GPU which is problematic for the current UI stack, Wayland helps here
 +different USB wifi devices failed horribly, demo using Intel PCIe, only one USB (Atheros) worked at all
 +why separate GPU - why not use virtual output from GPU? should be supported and let the GPU encode the stream? implementation is also providing userspace APIs to bypass decode/​encode
 +
 +what really is missing?
 + - supplicant that works - e.g. P2P_DEVICE support can cause spurious errors
 + - NetworkManager/​Connman support for P2P (the latter is WIP)
 +
 +
 +1:45 -- Extremely high speed and high coverage home wifi -- Avery Pennarun
 +2:00 -- Cloud-based wireless packet capture, sharing, analysis, and diagnosis -- Avery Pennarun
 +"How we're doing WiFi @Google Fiber"
 +problem of selling gbit service - wifi is too slow, WAN faster than LAN
 +no longer a question of whether customers will be disappointed or not, just a matter of how disappointed :)
 +wifi speeds drop with distance - so now APs in each TV box; this causes problems roaming
 +http://​6-dot-gfblip.appspot.com/​ - measures latency in http session, can do DNS queries
 +new version will record & upload the data for customer support purposes
 +other tool: isoping to measure upstream/​downstream latency/​jitter
 +http://​wavedroplet.appspot.com/​ - upload pcap (radiotap headers) and show 802.11 data graphed
 +q: bufferbloat,​ how does it affect you?
 +a: when things are going well the systems often can't even catch up with the wan link so not a problem there; solutions to bufferbloat all on ethernet so far but wifi is really what they need, where it's a much harder problem
 +...
 +network stack may have to be changed to deal with more dynamic queues due to changing stations/​link conditions
 +
 +2:15 -- Break
 +
 +2:30 -- AP Group Features -- Helmut Schaa
 +AP group is a set of APs managed centrally, a lot of problems could be solved with such a framework if it were to exist
 +proposal to create a new protocol specific to hostapd for cooperation between APs
 +
 +
 +2:45 -- "​Minstrel-Blues"​ - Practical Joint Rate und Power Control in Linux mac80211 with todays 802.11b/​g/​a/​n Chips -- Thomas Huehn
 +rate control algorithm that incorporates tx power adjustment to reduce interference
 +modifications to minstrel, working w/ ath5k & ath9k
 +new mac80211 driver flags for power control capabilities
 +current work to extend to minstrel_ht
 +
 +3:00 -- generalized firmware coredump support ​ -- Johannes Berg
 +framework merged in 3.18
 +register a buffer or method for access
 +creates device_coredump device class in sysfs
 +implemented in iwlwifi
 +not specific to wifi drivers
 +
 +3:15 -- Break
 +
 +3:30 -- randomized MAC address scanning -- Johannes Berg
 +originally from Apple (keybote), helps with privacy concerns
 +implementation in hostapd
 +When to randomize? ​ randomization breaks some use cases
 +only when not associated? ​ pre-association exchanges
 +should not be enabled by default -- end user option? ​ probably...
 +
 +3:45 -- Batman mesh -- Simon Wunderlich
 +overview of l2 mesh technology, generally used w/ wireless networking
 +mesh connection based on ad-hoc links
 +protocol decisions based on packet loss (OGMv2)
 +url: open-mesh.org
 +
 +4:00 -- 802.11s mesh -- Kathy Giori
 +new life for 802.11s, QCA acquired copyrights from Cozybit for firmware (could be used in ath10k)
 +some implementations can work better by breaking the spec slightly -- will we accept patches? ​ (likely yes, not guaranteed)
 +
 +4:15 -- regulatory -- Luis Rodriguez
 +efforts to push CRDA functionality into the kernel
 +timing issues exist w/ userland implementation,​ some use the in-kernel static database instead
 +use in-kernel signing infrastructure to validate rules database BLOBs coming from userland</​code>​
en/developers/summits/duesseldorf-2014.txt · Last modified: 2015/01/26 09:49 (external edit)