#3341 NORM First D: Failure to transfer large files over mesh interface while associated

[Zarro Boogs per Child]

#3341: Failure to transfer large files over mesh interface while associated
-----------------------+----------------------------------------------------
  Reporter:  jcardona  |       Owner:  ashish                           
      Type:  defect    |      Status:  new                              
  Priority:  normal    |   Milestone:  First Deployment, V1.0           
 Component:  wireless  |     Version:  Development build as of this date
Resolution:            |    Keywords:                                   
  Verified:  1         |  
-----------------------+----------------------------------------------------

Comment(by ashish):

 Wireless firmware release 5.110.19p5 allows large file, of the order of
 100MBs, copy through scp/sftp where receiver is scanning with a delay
 between successive scan of 0.5s, i.e., receiver is running the following
 command 'iwlist eth0 scan; sleep 0.5; done '.

 A continuous scan may delay route discovery as route discovery action
 frames
 cannot be transmisted if firmware is on channel other than the
 associated (AP)/joined (Ad-hoc)/mesh channel.
 When receiver is scanning, a data traffic from transmiter will generate
 multiple route discovery --this is due to the fact that while recevier is
 sending frame, transmitter might be on the other channel, and therefore,
 all
 MAC retries will be over and transmitter will assume broken link and start
 route discovery.
 Similarly, continous scan affects ARP discovery latency.

 scp/sftp have their own flow control on top of TCP. scp/sftp display
 'stalled'
 when they do not receive data for more than 5s. However, scp/sftp are
 robust
 enough to resume transfer afer a while.

 I do not know much about NetworkManager behavior, however, as per my
 practical experience observe that NetworkManager does not execute active
 scan too frequently. Therefore, I guess scp/sftp should work when
 NetworkManager is running and it has already discovered and joined the
 network.

-- 
Ticket URL: <https://dev.laptop.org/ticket/3341#comment:11>
One Laptop Per Child <https://dev.laptop.org>
OLPC bug tracking system