Date: Tue, 18 Apr 2017 08:28:17 -0500 (CDT)
From: Christoph Lameter <cl@linux.com>
To: Sergey Senozhatsky <sergey.senozhatsky.work@gmail.com>
cc: Minchan Kim <minchan@kernel.org>, Joonsoo Kim <iamjoonsoo.kim@lge.com>,
        Andrew Morton <akpm@linux-foundation.org>,
        Michal Hocko <mhocko@suse.com>, Vlastimil Babka <vbabka@suse.cz>,
        linux-mm@kvack.org, linux-kernel@vger.kernel.org, kernel-team@lge.com,
        Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Subject: Re: copy_page() on a kmalloc-ed page with DEBUG_SLAB enabled (was
 "zram: do not use copy_page with non-page alinged address")
In-Reply-To: <20170418104222.GB558@jagdpanzerIV.localdomain>
Message-ID: <alpine.DEB.2.20.1704180825460.13506@east.gentwo.org>
References: <20170417014803.GC518@jagdpanzerIV.localdomain> <alpine.DEB.2.20.1704171016550.28407@east.gentwo.org> <20170418104222.GB558@jagdpanzerIV.localdomain>
Content-Type: text/plain; charset=US-ASCII
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 966
Lines: 25

On Tue, 18 Apr 2017, Sergey Senozhatsky wrote:

> > Simple solution is to not allocate pages via the slab allocator but use
> > the page allocator for this. The page allocator provides proper alignment.
>
> sure, but at the same time it's not completely uncommon and unseen thing
>
> ~/_next$ git grep kmalloc | grep PAGE_SIZE | wc -l
> 75

Of course if you want a PAGE_SIZE object that is not really page aligned
etc then its definitely ok to use.

> not all, if any, of those pages get into copy_page(), of course. may be... hopefully.
> so may be a warning would make sense and save time some day. but up to MM
> people to decide.

Slab objects are copied using memcpy. copy_page is for pages aligned to
page boundaries and the arch code there may have additional expectations
that cannot be met by the slab allocators.

> p.s. Christoph, FYI, gmail automatically marked your message
>      as a spam message, for some reason.

Weird. Any more details as to why?