Re: kmalloc without GFP_xxx?

[Jens Axboe <axboe@suse.de>]

On Wed, Jun 29 2005, Denis Vlasenko wrote:
> On Wednesday 29 June 2005 14:15, Jens Axboe wrote:
> > On Wed, Jun 29 2005, Denis Vlasenko wrote:
> > > So why can't we have kmalloc_auto(size) which does GFP_KERNEL alloc
> > > if called from non-atomic context and GFP_ATOMIC one otherwise?
> > 
> > Because it's a lot better in generel if we force people to think about
> > what they are doing wrt memory allocations. You should know if you are
> > able to block or not, a lot of functions exported require you to have
> > this knowledge anyways. Adding these auto-detection type functions
> > encourages bad programming, imho.
> 
> Those 'bad programming' people can simply use GFP_ATOMIC always, no?
> This would be even worse because kmalloc_auto() will sleep
> if it's allowed, but GFP_ATOMIC would not.

Sure, you can't stop people from doing bad programming. But I don't
think we should aid them along the way.

-- 
Jens Axboe