[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20170401215118.GB31653@khazad-dum.debian.net>
Date: Sat, 1 Apr 2017 18:51:18 -0300
From: Henrique de Moraes Holschuh <hmh@....eng.br>
To: Wols Lists <antlists@...ngman.org.uk>
Cc: Ming Lei <tom.leiming@...il.com>, Arnd Bergmann <arnd@...db.de>,
Shaohua Li <shli@...nel.org>, NeilBrown <neilb@...e.com>,
Jens Axboe <axboe@...com>, "colyli@...e.de" <colyli@...e.de>,
Guoqing Jiang <gqjiang@...e.com>,
Mike Christie <mchristi@...hat.com>,
"open list:SOFTWARE RAID (Multiple Disks) SUPPORT"
<linux-raid@...r.kernel.org>,
Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] Revert "md: raid1: use bio helper in process_checks()"
On Tue, 28 Mar 2017, Wols Lists wrote:
> What Arnd is doing is commonly called "defensive programming", and
> unfortunately reality shows us that it is usually worth its weight in
> gold. That's why you put ASSERTs in code - so that if somebody does
> something stupid by accident, it blows up. This is just more of the same.
You know, BUG_ON(vcnt1 != vcnt2) might address this quite nicely [if you
want to go the assert() way, that is!], since __attribute__((noreturn))
is set for the codepath taken when the BUG_ON condition triggers...
That said, if there is something less insane to be done than OOPsing
when the world has gone strange and vcnt1 != vcnt2, that ought to be a
better solution...
All of this assumes vcnt1 == vcnt2 is really the only possibily correct
situation.
--
Henrique Holschuh
Powered by blists - more mailing lists