linux-kernel - Re: [PATCH] slab.h: Avoid using & for logical and of booleans

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: <51d8ac56-d4f6-4906-92a8-18b5fb42b198@suse.cz>
Date:   Fri, 9 Nov 2018 22:31:48 +0100
From:   Vlastimil Babka <vbabka@...e.cz>
To:     "Darryl T. Agostinelli" <dagostinelli@...il.com>
Cc:     Andrew Morton <akpm@...ux-foundation.org>,
        David Laight <David.Laight@...lab.com>,
        'Bart Van Assche' <bvanassche@....org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        Mel Gorman <mgorman@...hsingularity.net>,
        Christoph Lameter <cl@...ux.com>, Roman Gushchin <guro@...com>
Subject: Re: [PATCH] slab.h: Avoid using & for logical and of booleans

On 11/9/18 8:47 PM, Darryl T. Agostinelli wrote:
> On Fri, Nov 09, 2018 at 08:16:07PM +0100, Vlastimil Babka wrote:
>> On 11/9/18 8:00 PM, Andrew Morton wrote:
>>> On Fri, 9 Nov 2018 09:12:09 +0100 Vlastimil Babka <vbabka@...e.cz> wrote:
>>>
>>>> Multiple people have reported the following sparse warning:
>>>>
>>>> ./include/linux/slab.h:332:43: warning: dubious: x & !y
>>>>
>>>> The minimal fix would be to change the logical & to boolean &&, which emits the
>>>> same code, but Andrew has suggested that the branch-avoiding tricks are maybe
>>>> not worthwile. David Laight provided a nice comparison of disassembly of
>>>> multiple variants, which shows that the current version produces a 4 deep
>>>> dependency chain, and fixing the sparse warning by changing logical and to
>>>> multiplication emits an IMUL, making it even more expensive.
>>>>
>>>> The code as rewritten by this patch yielded the best disassembly, with a single
>>>> predictable branch for the most common case, and a ternary operator for the
>>>> rest, which gcc seems to compile without a branch or cmov by itself.
>>>>
>>>> The result should be more readable, without a sparse warning and probably also
>>>> faster for the common case.
>>>>
>>>> Reported-by: Bart Van Assche <bvanassche@....org>
>>>> Reported-by: Darryl T. Agostinelli <dagostinelli@...il.com>
>>>> Suggested-by: Andrew Morton <akpm@...ux-foundation.org>
>>>> Suggested-by: David Laight <David.Laight@...LAB.COM>
>>>> Fixes: 1291523f2c1d ("mm, slab/slub: introduce kmalloc-reclaimable caches")
>>>> Signed-off-by: Vlastimil Babka <vbabka@...e.cz>
>>>> ---
>>>>  include/linux/slab.h | 24 ++++++++++++------------
>>>>  1 file changed, 12 insertions(+), 12 deletions(-)
>>>>
>>>> diff --git a/include/linux/slab.h b/include/linux/slab.h
>>>> index 918f374e7156..18c6920c2803 100644
>>>> --- a/include/linux/slab.h
>>>> +++ b/include/linux/slab.h
>>>> @@ -304,6 +304,8 @@ enum kmalloc_cache_type {
>>>>  	KMALLOC_RECLAIM,
>>>>  #ifdef CONFIG_ZONE_DMA
>>>>  	KMALLOC_DMA,
>>>> +#else
>>>> +	KMALLOC_DMA = KMALLOC_NORMAL,
>>>>  #endif
>>>>  	NR_KMALLOC_TYPES
>>>>  };
>>>
>>> I don't think this works correctly.  Resetting KMALLOC_DMA to 0 will
>>> cause NR_KMALLOC_TYPES to have value 1.
>>
>> Doh, right! Thanks for catching this.
>>
>> This? Not terribly elegant, but I don't see a nicer way right now...
>>
> 
> How about the solution I proposed yesterday? 
> 
> https://lkml.org/lkml/2018/11/9/750
> 
> It doesn't involve any tricks. 

It doesn't remove the "trick" that calculates return value as a sum of
booleans multiplying constants. The patch converts one part of the
expression of those booleans to a ternary operator. I think the result
is even harder to follow and meanwhile Andrew's suggestion was to remove
all the tricks.

> As it is, this sparse warning is begging for a trick. Let's not 
> oblidge it to much.

The sparse warning could be silenced just by changing '&' to '&&' which
would emit the same code. But we decided to untrick the code.