Re: [PATCH][GIT PULL][for 2.6.35] tracing: Add alignment tosyscall metadata declarations

[Sam Ravnborg]

On Fri, Jul 09, 2010 at 05:05:50PM -0400, Steven Rostedt wrote:
> On Fri, 2010-07-09 at 22:53 +0200, Sam Ravnborg wrote:
> 
> > But in this case we have no control of the value of "." (current address)
> > when we have processed (_ftrace_events) so it may even be at a 2 byte boundary.
> > The linker will add padding as needed to satisfy the alignmnet of
> > __syscalls_metadata - but that padding will be inbetween "." and the first
> > member in __syscalls_metadata.
> 
> Fine, but this is a separate issue.
No - this is the exact issue we are facing here.

In the linker script we have no control what-so-ever on the alignment of "."
so relying on any alignment is a potential bug.

In this particular case gcc decide to increase the alignment of
syscall_metadata from 8 to 16.

And then suddenly things fall apart because we rely on "." being
properly aligned - and it is only 4 byte aligned.

So you fixed this by forcing gcc to use a smaller alignment
for syscall_metadata (from 16 => 4).
But what happens next time when "." is two-byte aligned?
We will not change aligment down to 2...

So to fix this properly we need to make sure that "." is properly
aligned to the alignment of syscall_metadata.

In other words - the linker script needs to be fixed too.
But the fix in the .h file is required so we know the alignmnet
of syscall_metadata.

	Sam
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/