Re: [PATCH 0/2] jump label: update for .39

[David Daney <ddaney@caviumnetworks.com>]

On 03/10/2011 10:35 AM, Steven Rostedt wrote:
> On Thu, 2011-03-10 at 13:20 -0500, Jason Baron wrote:
>> On Thu, Mar 10, 2011 at 01:04:01PM -0500, Steven Rostedt wrote:
>> Now its up to the linker to decide where to place that element. Can we
>>> guarantee that it will always be on an 8 byte boundery?
>>> Hmm, I wonder if we could add a .ALIGN sizeof(long) before that?
>>>
>>
>> we can. Sparc does it, see: arch/sparc/include/asm/jump_label.h.
>>
>> So I guess it would be .align 8 for 64-bit and .align 4 for 32-bit...
>
> Now what about the vmlinux.ld? That has a align 8. Is that a one time
> short (align to 8 bytes here), or will it carry through aligning the
> rest of the section. If not, perhaps we may be OK.
>

That is the alignment of the start of the section in the *output* 
object.  The linker will respect the requested alignment of the various 
__jump_table sections in the *input* objects.  So adding an .align 8 
before emitting data into the __jump_table section would not hurt (It 
could even help.)

I empirically determined that if no .align is specified, the requested 
alignment defaults to 4 for x86_64.

The alignment requested by the assembler will have to satisfy *all* the 
requested alignments, so manually forcing everything to .align 8 (or 
.align 4 for 32-bit) should ensure that the linker doesn't put in any holes.

David Daney