Re: [PATCH] get_nodes mask miscalculation

[Andi Kleen <ak@muc.de>]

On Tue, Aug 10, 2004 at 09:54:08AM -0500, Brent Casavant wrote:
> On Tue, 10 Aug 2004, Andi Kleen wrote:
> 
> > The idea behind this is was to make it behave like select.
> > And select passes highest valid value + 1.
> 
> Not that my opinion is worth much, but this seems a very peculiar
> interface to mimic.

It is the only system call I am aware of that passes a variable
length bitmap.  Do you know of any others? 

> > nodemask."
> 
> Actually, I worked from the following description:
> 
> mbind(3)
> 	. . .
> 	nodemask is a bit field of nodes that contains up to
> 	maxnode bits.
> 	. . .
> 
> This very clearly indicates that I should pass the number of bits
> in the nodemask field, not the number of bits plus one.

I really need to rewrite the mbind manpage - it was a quick
hack job and has several other errors.

> > Problem is that you'll break all existing libnuma binaries
> > which pass NUMA_MAX_NODES + 1. In your scheme the kernel
> > will access one bit beyond the bitmap that got passed,
> > and depending on its random value you may get a failure or not.
> 
> Well, again, not that my opinion carries much weight (nor really should
> it), but this whole off-by-one (or two) interface seems unnecessarily
> confusing.  Is there no way we can get this fixed?  The temporary pain

Sure, new system call slots would do with a small compat wrapper
for the old calls. I'm just not sure it is worth it. I personally
don't find it that confusing, but I'm probably not a good benchmark
for this. 

> of breaking the relatively new libnuma seems worth the price of getting
> this forever cleaned up.

I don't think breaking existing binaries is a good idea. Linux
has higher standards. 

-Andi
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"aart@kvack.org"> aart@kvack.org </a>