linux-kernel - Re: Slab infrastructure for bulk object allocation and freeing V2

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-Id: <20150402134239.8e8c538103640d697246ba6a@linux-foundation.org>
Date:	Thu, 2 Apr 2015 13:42:39 -0700
From:	Andrew Morton <akpm@...ux-foundation.org>
To:	Christoph Lameter <cl@...ux.com>
Cc:	Jesper Dangaard Brouer <brouer@...hat.com>,
	linux-kernel@...r.kernel.org, linux-mm@...ck.org,
	akpm@...uxfoundation.org, Pekka Enberg <penberg@...nel.org>,
	iamjoonsoo@....com
Subject: Re: Slab infrastructure for bulk object allocation and freeing V2

On Thu, 2 Apr 2015 09:25:37 -0500 (CDT) Christoph Lameter <cl@...ux.com> wrote:

> > What's the reason for returning a partial result when ENOMEM?  Some
> > callers will throw away the partial result and simply fail out.  If a
> > caller attempts to go ahead and use the partial result then great, but
> > you can bet that nobody will actually runtime test this situation, so
> > the interface is an invitation for us to release partially-tested code
> > into the wild.
> 
> Just rely on the fact that small allocations never fail? The caller get
> all the requested objects if the function returns?

I'd suggest the latter: either the callee successfully allocates all
the requested objects or it fails.

> > Instead of the above, did you consider doing
> >
> > int __weak kmem_cache_alloc_array(struct kmem_cache *s, gfp_t flags, size_t nr,
> >
> > ?
> >
> > This way we save a level of function call and all that wrapper code in
> > the allocators simply disappears.
> 
> I think we will need the auxiliary function in the common code later
> because that allows the allocations to only do the allocations that
> can be optimized and for the rest just fall back to the generic
> implementations. There may be situations in which the optimizations wont
> work. For SLUB this may be the case f.e. if debug options are enabled.

hm, OK.  The per-allocator wrappers could be made static inline in .h
if that makes sense.


With the current code, gcc should be able to convert the call into a
tailcall.

<checks>

nope.

kmem_cache_free_array:
	pushq	%rbp	#
	movq	%rsp, %rbp	#,
	call	__kmem_cache_free_array	#
	leave
	ret

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