#12781 HIGH 13.2.0-: XO-4 Android performance regression on 1GB units

Zarro Boogs per Child bugtracker at laptop.org
Tue May 20 21:51:44 EDT 2014


#12781: XO-4 Android performance regression on 1GB units
--------------------------------+-------------------------------------------
           Reporter:  Quozl     |       Owner:  Quozl                            
               Type:  defect    |      Status:  new                              
           Priority:  high      |   Milestone:  13.2.0-android                   
          Component:  android   |     Version:  Development build as of this date
         Resolution:            |    Keywords:                                   
        Next_action:  diagnose  |    Verified:  0                                
Deployment_affected:            |   Blockedby:                                   
           Blocking:            |  
--------------------------------+-------------------------------------------

Comment(by Quozl):

 A theory proposed by Ben was that the depletion of HighFree in
 /proc/meminfo may be a trigger for the problem.

 Adopting the upstream sanity_check_meminfo implementation in
 arch/arm/mm/mmu.c and then forcing the low to high boundary to 512MB
 (0x20000000) showed correct output in /proc/zoneinfo, HighFree no longer
 depleted in /proc/meminfo, but continued alloc_contig_range errors.

 So the problem is not a shortage of memory.

 Removing several of the repeating errors by patching them to be either
 rate limited or fire once only, made the system more responsive.

 The remaining delay appears to be related to the persistent loading of the
 bluetooth module, which fails to load because of the kernel version string
 change.

  * [http://dev.laptop.org/~quozl/y/1Wmvd7.txt accumulated patches] used in
 diagnosis

-- 
Ticket URL: <http://dev.laptop.org/ticket/12781#comment:2>
One Laptop per Child <http://laptop.org/>
One Laptop per Child bug tracking system


More information about the Bugs mailing list