Re: Bug#609371: linux-image-2.6.37-trunk-sparc64: module scsi_mod:

[Jesper Nilsson <jesper.nilsson@axis.com>]

On Mon, Jan 17, 2011 at 07:07:55AM +0100, David Miller wrote:
> From: David Miller <davem@davemloft.net>
> Date: Sat, 15 Jan 2011 21:17:22 -0800 (PST)
> 
> [ Please, everyone, retain the full CC: on all replies, thanks.  Some
>   people are replying only into the debian bug alias, and that loses
>   information and exposure for fixing this bug.  ]
> 
> > I think the problem we have here is that the _ftrace_events section is
> > not aligned sufficiently.  That ".align 4" mnemonic is a good indication
> > of this.  It should at least "8" on sparc64.
> 
> I did some more research.
> 
> Although I've seen commentary to the contrary, in fact using a too-small
> __attribute__((aligned())) directive will lower the alignment of data
> members, and yes that means it will lower the alignemnt to be below the
> natural and required alignment for the given type.
> 
> So if you have, on 64-bit:
> 
> struct foo {
> 	void *bar;
> };
> 
> static struct foo test __attribute__((__aligned__(4)));
> 
> The compiler will emit "test" with 4-byte alignment into the data
> section, even though 8-byte alignment is required for "test.bar"
> 
> Assuming we wanted that to actually happen, the GCC manual is very
> explicit to state that in order for this to work, such down-aligned
> data structures must also use the "packed" attribute.
> 
> I think we want none of this, and I think we should elide the align
> directives entirely, or at least fix them so we don't get unaligned
> stuff on 64-bit.
> 
> Ugh, and I just noticed that include/linux/klist.h does this fixed
> alignment of "4" too, where is this stuff coming from?  It's
> wrong on 64-bit, at best.  But I can't see the impetus behind doing
> this at all in the first place.
> 
> Oh, this is some CRIS thing, because it only byte aligns.  See:
> 
> commit c0e69a5bbc6fc74184aa043aadb9a53bc58f953b
> Author: Jesper Nilsson <Jesper.Nilsson@axis.com>
> Date:   Wed Jan 14 11:19:08 2009 +0100
> 
>     klist.c: bit 0 in pointer can't be used as flag
> 
> That's where the klist one comes from.

Yup, this one could instead be solved by introducing a "flags" field
in the struct, but that was considered a too large impact fix.

> The ftrace ones come from:
> 
> commit 86c38a31aa7f2dd6e74a262710bf8ebf7455acc5
> Author: Jeff Mahoney <jeffm@suse.com>
> Date:   Wed Feb 24 13:59:23 2010 -0500
> 
>     tracing: Fix ftrace_event_call alignment for use with gcc 4.5
>     
> We really can't handle this that way, it's going to break stuff
> on 64-bit systems at the very least.
> 
> How about we use __BIGGEST_ALIGNMENT__ or something arch-defined value
> instead?

From CRIS-standpoint that would be fine.

/^JN - Jesper Nilsson
-- 
               Jesper Nilsson -- jesper.nilsson@axis.com