Re: [dm-devel] edits for r16 of shared snapshot patches [was: Re: userspace patches for shared snapshots]

[Mike Snitzer <snitzer redhat com>]

On Tue, Mar 09 2010 at  3:41am -0500,
Mikulas Patocka <mpatocka redhat com> wrote:

> > You can see the edits here:
> > http://people.redhat.com/msnitzer/patches/multisnap/kernel/2.6.33/r16a/r16_edits.patch
> > 
> > Boils down to:
> > * use __func__ rather than hardcoding the function name
> >   - this fixed ~3 inconsistencies (incorrect function names) and should
> >     help if/when we do any function renaming in later phases of review
> 
> It consumes one more stack word for every function where this conversion 
> was performed ... I don't know if it's worth it ... likely it doesn't 
> matter.
> 
> The problem here is that GCC is preallocating the space for all outgoing 
> arguments in the function prologue, so if the function has something like 
> this
> if (bug_happened) {
> 	printk("%s: bug happened: %d %d %d %d %d %d %d %d", __func__, a, 
> b, c, d, e, f, g, h, i);
> }
> the whole function always wastes 10 stack words, even if the bug doesn't 
> happen :(
> 
> But that one word because of __func__ maybe doesn't matter (there are much 
> more wasted already in these printks). Or do you think that it's better to 
> not do this __func__ conversion?

I think the use of __func__ reduces the maintenance cost of keeping the
printk updated if the code were to change.  Like I said, I found ~3
remaining function name inconsistencies.  Plus I had already fixed a few
others in my previous large whitespace patch.

Anyway, I understand your point on the extra word associated with using
__func__ but feel using it is more helpful.

Mike