#9557 NORM 1.5-har: os33 - internal SD - IO error

[Zarro Boogs per Child]

#9557: os33 - internal SD - IO error
--------------------------------+-------------------------------------------
           Reporter:  mikus     |       Owner:  wad                              
               Type:  defect    |      Status:  closed                           
           Priority:  normal    |   Milestone:  1.5-hardware-C                   
          Component:  hardware  |     Version:  Development build as of this date
         Resolution:  invalid   |    Keywords:  os33 q3a14                       
        Next_action:  diagnose  |    Verified:  0                                
Deployment_affected:            |   Blockedby:                                   
           Blocking:            |  
--------------------------------+-------------------------------------------

Comment(by Quozl):

 Indeed there were many ext4 problems in the kernel we were using at the
 time ... given the huge list of fixes made by the kernel point releases
 since.

 Looking again at the messages file, the failure was an attempt to read
 block 7189535 or 7189511.  The read was a multi-sector request, which is
 why there were multiple errors for each event as the kernel tried smaller
 requests.

 Considering whether the read failed because it was beyond the end of the
 device (since this is a typical ext3 or ext4 filesystem metadata
 corruption scenario):

  * the dmesg shows "mmcblk0: mmc0:aaaa SU04G 3.69 GiB"
  * my dmesg shows the same,
  * my /proc/partitions shows 3872256 blocks, therefore 7744512 sectors.

 The read request was within the total size of the device.

 My gut feel is not filesystem bugs, but rather that the SD card could not
 read some data and was telling the kernel this.  The flagged bad sectors
 in the device would have since been cleared by fs-update.

 Cause of the bad sectors might be power fail during write.

 An !OpenFirmware recipe that reads the whole device would have been handy.

-- 
Ticket URL: <http://dev.laptop.org/ticket/9557#comment:6>
One Laptop Per Child <http://laptop.org/>
OLPC bug tracking system