#9975 BLOC 1.5-sof: bash seg faults when running suspend/resume tests

[Zarro Boogs per Child]

#9975: bash seg faults when running suspend/resume tests
--------------------------------+-------------------------------------------
           Reporter:  wad       |       Owner:  pgf                               
               Type:  defect    |      Status:  new                               
           Priority:  blocker   |   Milestone:  1.5-software-update               
          Component:  distro    |     Version:  1.5 Software Build os64 aka 10.1.0
         Resolution:            |    Keywords:                                    
        Next_action:  diagnose  |    Verified:  0                                 
Deployment_affected:            |   Blockedby:                                    
           Blocking:            |  
--------------------------------+-------------------------------------------

Comment(by Quozl):

 Having a rethink because this is still blocker.

 I'm worried this may be memory related, and it might not be possible to
 get a core dump.

 On another system in my house there were several, regular and reproducible
 [http://pastebin.ca/1786731 segfaults reported by kernel] that I fixed
 yesterday ... a stick of RAM had gone bad, at a particular address, as
 identified by memtest.

 I'm not able to conclude that this ticket is certainly memory related, but
 there is a correlation.

 The message reported in the ticket description is from
 arch/x86/kernel/traps.c probably in the function do_general_protection().

 Firstly, if the problem happens frequently, there is chance that
 printk_ratelimit() will have prevented duplicates.  We know there was one.
 We don't know there wasn't more than one.

 Secondly, this is happening for an unhandled signal ... which for an
 ordinary segmentation fault caused by programming error doesn't tend to
 happen ... you don't get kernel messages for normal segmentation faults,
 in my experience.  This would be testable.

 Thirdly, the numbers, such as the instruction pointer and stack pointer,
 seem reasonable, but might benefit from a bit more analysis.  I lack clue
 there.

-- 
Ticket URL: <http://dev.laptop.org/ticket/9975#comment:11>
One Laptop Per Child <http://laptop.org/>
OLPC bug tracking system