linux-kernel - Re: number of elements or bytes for kfifo

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] [day] [month] [year] [list]

Message-ID: <1282200778.3752.7.camel@wall-e.seibold.net>
Date:	Thu, 19 Aug 2010 08:52:58 +0200
From:	Stefani Seibold <stefani@...bold.net>
To:	Huang Ying <ying.huang@...el.com>
Cc:	Andrew Morton <akpm@...ux-foundation.org>,
	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: number of elements or bytes for kfifo_in etc

Am Donnerstag, den 19.08.2010, 09:15 +0800 schrieb Huang Ying:
> On Thu, 2010-08-19 at 08:32 +0800, Huang Ying wrote:
> > Hi, Stefani,
> > 
> > On Thu, 2010-08-19 at 00:35 +0800, Stefani Seibold wrote:
> > > Am Mittwoch, den 18.08.2010, 13:29 +0800 schrieb Huang Ying:
> > > > Hi, Stefani,
> > > > 
> > > > In new kfifo implementation in 2.6.36-rc1, description of the third
> > > > parameter is:
> > > > 
> > > > "@n: number of elements to be added"
> > > > 
> > > > But if my understanding is correct, the actual implementation is:
> > > > 
> > > > "@n: number of bytes to be added"
> > > > 
> > > 
> > > Number of bytes is wrong, this is only valid for byte stream fifo's
> > > where an element is a byte.
> > > 
> > > Have a look at samples/kfifo/inttype-example.c where an element is a int
> > > type.
> > 
> > Yes. You are right. I misunderstood your code. Sorry for bothering.
> 
> And I found __kfifo->esize is used in __kfifo_in/out etc to convert
> between elements and bytes. Maybe we can pull the conversion code to
> corresponding macro, because where esize is constant and may be
> optimized to bit shifting or nothing. What do you think about that?
> 

esize isn't constant, especially from the point of view inside
__kfifo_in and so on. The is ony good reason not to pass the size of the
element using sizeof(*fifo->type) as a parameter, it will generate more
code. Storing the size of the element in the kfifo structure is the
better decision.

But if you found a better way and can prove it under any circumstance
with a list of measurement results u will get an ack!

- Stefani


--
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/