On Wed, Jun 16, 2010 at 10:47 PM, Neil Brown [off-list ref] wrote:

On Mon, 14 Jun 2010 12:47:23 +0200
Nagilum [off-list ref] wrote:

quoted

----- Message from neilb@suse.de ---------
     Date: Mon, 14 Jun 2010 09:15:18 +1000
     From: Neil Brown [off-list ref]
  Subject: Re: Problem re-shaping RAID6
       To: Jérôme Poulin [off-list ref]
       Cc: linux-raid [off-list ref]

quoted

Thanks.
I fixed this bug a slightly different way

-               blocks = ochunk/512 * nchunk/512 * odata * ndata / a;
+               blocks = (ochunk/512) * (nchunk/512) * odata * ndata / a;


See
http://neil.brown.name/git?p=mdadm;a=commitdiff;h=200871adf9e15d5ad985f28c349fd89c386ef48a

Those static numbers always make my nose wrinkle.
Don't we have the blocksize somewhere already? I'm also concerned what
happens when true 4k sectors are used..

A sector will always be 512 bytes to Linux, even when we have drives that can
only do IO in multiples of 8 sectors.  Changing that would cause way to many
headaches.

Yes,  I could possibly use a define for '512'.  Some times that is
appropriate, but I thing 512 is so clearly "bytes_per_sector" it would just
add an unnecessary level of indirection.
It is a question of taste really - no right answers.

Thanks,
NeilBrown

--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Slightly off topic, but wouldn't an 'always' like that be a valid
reason for starting a 2.7 or 3.0 branch; that is something where major
sections of code would /have/ to be changed even if interfaces for
other sections might remain static?
--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html