linux-kernel - Re: [PATCH] SLUB: Do not fallback to mininum order if __GFP

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]

Date:   Wed, 18 Apr 2018 11:49:22 -0700 (PDT)
From:   David Rientjes <rientjes@...gle.com>
To:     Mikulas Patocka <mpatocka@...hat.com>
cc:     Christopher Lameter <cl@...ux.com>,
        Vlastimil Babka <vbabka@...e.cz>,
        Mike Snitzer <snitzer@...hat.com>,
        Matthew Wilcox <willy@...radead.org>,
        Pekka Enberg <penberg@...nel.org>, linux-mm@...ck.org,
        dm-devel@...hat.com, Joonsoo Kim <iamjoonsoo.kim@....com>,
        Andrew Morton <akpm@...ux-foundation.org>,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH] SLUB: Do not fallback to mininum order if __GFP_NORETRY
 is set

On Wed, 18 Apr 2018, Mikulas Patocka wrote:

> > Mikulas Patoka wants to ensure that no fallback to lower order happens. I
> > think __GFP_NORETRY should work correctly in that case too and not fall
> > back.
> > 
> > 
> > 
> > Allocating at a smaller order is a retry operation and should not
> > be attempted.
> > 
> > If the caller does not want retries then respect that.
> > 
> > GFP_NORETRY allows callers to ensure that only maximum order
> > allocations are attempted.
> > 
> > Signed-off-by: Christoph Lameter <cl@...ux.com>
> > 
> > Index: linux/mm/slub.c
> > ===================================================================
> > --- linux.orig/mm/slub.c
> > +++ linux/mm/slub.c
> > @@ -1598,7 +1598,7 @@ static struct page *allocate_slab(struct
> >  		alloc_gfp = (alloc_gfp | __GFP_NOMEMALLOC) & ~(__GFP_RECLAIM|__GFP_NOFAIL);
> > 
> >  	page = alloc_slab_page(s, alloc_gfp, node, oo);
> > -	if (unlikely(!page)) {
> > +	if (unlikely(!page) && !(flags & __GFP_NORETRY)) {
> >  		oo = s->min;
> >  		alloc_gfp = flags;
> >  		/*
> 
> No, this would hit NULL pointer dereference if page is NULL and 
> __GFP_NORETRY is set. You want this:
> 
> ---
>  mm/slub.c |    2 ++
>  1 file changed, 2 insertions(+)
> 
> Index: linux-2.6/mm/slub.c
> ===================================================================
> --- linux-2.6.orig/mm/slub.c	2018-04-17 20:58:23.000000000 +0200
> +++ linux-2.6/mm/slub.c	2018-04-18 17:04:01.000000000 +0200
> @@ -1599,6 +1599,8 @@ static struct page *allocate_slab(struct
>  
>  	page = alloc_slab_page(s, alloc_gfp, node, oo);
>  	if (unlikely(!page)) {
> +		if (flags & __GFP_NORETRY)
> +			goto out;
>  		oo = s->min;
>  		alloc_gfp = flags;
>  		/*
> 

I don't see the connection between the max order, which can be influenced 
by userspace with slub_min_objects, slub_min_order, etc, and specifying 
__GFP_NORETRY which means try to reclaim and free memory but don't loop.

If I force a slab cache to try a max order of 9 for hugepages as a best 
effort, why does __GFP_NORETRY suddenly mean I won't fallback to 
oo_order(s->min)?