linux-kernel - Re: [PATCH v2 1/3] mtd: introduce the mtd_pairing

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: <20160901181524.GA22366@localhost>
Date:   Thu, 1 Sep 2016 11:15:24 -0700
From:   Brian Norris <computersforpeace@...il.com>
To:     Boris Brezillon <boris.brezillon@...e-electrons.com>
Cc:     David Woodhouse <dwmw2@...radead.org>,
        linux-mtd@...ts.infradead.org, Richard Weinberger <richard@....at>,
        George Spelvin <linux@...encehorizons.net>,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH v2 1/3] mtd: introduce the mtd_pairing_scheme concept

Hi,

I've had this on my plate to respond to for a while now, and I haven't
brought myself to actually care that much about the choice. So I'll
respond now to keep from leaving you hanging, but I'm not sure I'm that
helpful :(

On Tue, Aug 09, 2016 at 12:42:18AM +0200, Boris Brezillon wrote:
> On Thu, 4 Aug 2016 12:37:51 +0800
> Brian Norris <computersforpeace@...il.com> wrote:
> > On Mon, Jun 20, 2016 at 03:50:16PM +0200, Boris Brezillon wrote:
> 
> > 
> > > + * (3 bits in a single cell). A pair should regroup all pages that are sharing
> > > + * the same cell. Pairs are then indexed in ascending order.
> > > + *
> > > + * @group is defining the position of a page in a given pair. It can also be
> > > + * seen as the bit position in the cell: page attached to bit 0 belongs to
> > > + * group 0, page attached to bit 1 belongs to group 1, etc.
> > > + *
> > > + * Example:
> > > + * The H27UCG8T2BTR-BC datasheet describes the following pairing scheme:
> > > + *
> > > + *		group-0		group-1
> > > + *
> > > + *  pair-0	page-0		page-4
> > > + *  pair-1	page-1		page-5
> > > + *  pair-2	page-2		page-8
> > > + *  ...
> > > + *  pair-127	page-251	page-255
> > > + *
> > > + *
> > > + * Note that the "group" and "pair" terms were extracted from Samsung and
> > > + * Hynix datasheets, and might be referenced under other names in other
> > > + * datasheets (Micron is describing this concept as "shared pages").  
> > 
> > Very, very helpful (to me, even though I'm moderately familiar with the
> > concepts, but hopefully moreso for others who want to read and
> > understand this). Thanks for writing this up.
> 
> Actually, the more I think about it, the more I doubt those terms are
> appropriate (even if they are widely used in technical documents).
> 
> How about using the following names instead:
> 
> struct mtd_cell_sharing_scheme {
> 	...
> };
> 
> struct mtd_cell_sharing_info {
> 	/* the bit position in the cell */
> 	int bitpos;
> 	/*
> 	 * What was previously known as 'pair': an id representing a

Wait, so you're replacing the literature's "pair" term with "group", but
the literature already used "group" to mean something else? That seems
to be an unwise choice. (Or I'm misreading you.)

> 	 * group of cells forming a 'pair of pages'.
> 	 * I can't find a good description/word for this concept. Do
> 	 * you have better ideas?
> 	 */
> 	int group;
> };
> 
> What do you think?

I think there's something to be said for matching the literature out
there, and I personally thought that simply providing a little bit of
clarifying explanation in the comments was sufficient. But if you feel
like choosing a more generic name is better, then that's probably OK
too. So other than the above comment (don't overload terms too freely!),
I'd use your judgment.

FWIW, it still takes me a while to parse what the "pair" and "group" (or
"bitpos" and "group" -- although "bitpos" is actually quite clear, so I
guess I like that) actually mean, so I tend to refer back to these
comments every time I'm reading it.

Brian