[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <20130627152335.c3a4c9f4c647cf4a2b263479@linux-foundation.org>
Date: Thu, 27 Jun 2013 15:23:35 -0700
From: Andrew Morton <akpm@...ux-foundation.org>
To: Ralf Baechle <ralf@...ux-mips.org>
Cc: Veli-Pekka Peltola <veli-pekka.peltola@...egiga.com>,
linux-kernel@...r.kernel.org, linux-arm-kernel@...ts.infradead.org,
linux-mips@...ux-mips.org, Russell King <linux@....linux.org.uk>,
Thomas Gleixner <tglx@...utronix.de>,
Ingo Molnar <mingo@...hat.com>,
"H. Peter Anvin" <hpa@...or.com>, x86@...nel.org,
Peter Zijlstra <peterz@...radead.org>,
Rusty Russell <rusty@...tcorp.com.au>
Subject: Re: [PATCH v2] mm: module_alloc: check if size is 0
On Thu, 27 Jun 2013 11:39:17 +0200 Ralf Baechle <ralf@...ux-mips.org> wrote:
> Imho de7d2b567d040e3b67fe7121945982f14343213d [mm/vmalloc.c: report more
> vmalloc failures] is overly strict in that it also reports zero-sized
> allocations. I consider such allocations stupid but legitimiate and often
> better preferrable over having to scatter checks for zero size all over
> place. So maybe something like below patch?
>
> ...
>
> --- a/mm/vmalloc.c
> +++ b/mm/vmalloc.c
> @@ -1679,7 +1679,10 @@ void *__vmalloc_node_range(unsigned long size, unsigned long align,
> unsigned long real_size = size;
>
> size = PAGE_ALIGN(size);
> - if (!size || (size >> PAGE_SHIFT) > totalram_pages)
> + if (unlikely(!size))
> + return NULL;
> +
> + if ((size >> PAGE_SHIFT) > totalram_pages)
> goto fail;
>
> area = __get_vm_area_node(size, align, VM_ALLOC | VM_UNLIST,
> @@ -1711,6 +1714,7 @@ fail:
> warn_alloc_failed(gfp_mask, 0,
> "vmalloc: allocation failure: %lu bytes\n",
> real_size);
> +
> return NULL;
> }
If the caller actually dereferences the returned pointer the kernel
will go oops, which should provide adequate notification of a
programming error ;) But all callers should be checking the return
value. So I worry about the by-far-most-common case where code does
size = some_screwed_up_calculation();
p = vmalloc(size);
if (!p)
return -ENOMEM;
So the mistake gets propagated back to who-knows-where as memory
exhaustion and thereby becomes a lot harder to diagnose.
How many callsites really truly need to be edited to avoid the warning?
Veli-Pekka's original patch would be neater if we were to add a new
void *__vmalloc_node_range_zero_size_ok(<args>)
{
if (size == 0)
return NULL;
return __vmalloc_node_range(<args>);
}
(with a better name than __vmalloc_node_range_zero_size_ok!)
--
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/
Powered by blists - more mailing lists