#12727 NORM Not Tri: OFW doesn't follow x86 linux boot protocol

[Zarro Boogs per Child]

#12727: OFW doesn't follow x86 linux boot protocol
---------------------------------+------------------------------------------
 Reporter:  dsd                  |                 Owner:  Quozl        
     Type:  defect               |                Status:  new          
 Priority:  normal               |             Milestone:  Not Triaged  
Component:  ofw - open firmware  |               Version:  not specified
 Keywords:                       |           Next_action:  never set    
 Verified:  0                    |   Deployment_affected:               
Blockedby:                       |              Blocking:               
---------------------------------+------------------------------------------
 I think OFW doesn't quite follow the steps described by the x86 boot
 protocol:
 http://lxr.linux.no/linux+v3.10/Documentation/x86/boot.txt#L1021

 New kernels (3.9 onwards) are detecting this through placing a 0xff
 sentinel in the zero page at address 0x1ef. That falls outside of the
 range of the data that should be loaded from the kernel image, instead it
 is part of the memory region that should have been zeroed out.

 The kernel boots and sees the sentinel value still in place and is not
 100% happy.
 http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=5dcd14ecd41ea2b3ae3295a9b30d98769d52165f

 I think these lines in cpu/x86/pc/linux.fth are responsible:

 {{{
 : code16-size  ( -- #bytes )   load-base h# 1f1 + c@ 1+  d# 512 *   ;

    load-base  0 +lp  code16-size  move   \ Copy the 16-bit stuff
    loaded code16-size /string  linux-base  swap  move  \ Copy the 32-bit
 stuff
 }}}

 But I can't quite get my head around them. Given that the 2nd line above
 is effectively "copy memory from start of the loaded kernel image to the
 place where the boot parameters will go" I see that as a suspect - but I
 can't quite get my head around code16-size logic or what the final line
 does.

-- 
Ticket URL: <http://dev.laptop.org/ticket/12727>
One Laptop Per Child <http://laptop.org/>
OLPC bug tracking system