x86: Slightly tweak the access_ok() C variant for better code
authorH. Peter Anvin <hpa@zytor.com>
Sat, 28 Dec 2013 00:52:47 +0000 (16:52 -0800)
committerH. Peter Anvin <hpa@zytor.com>
Sat, 28 Dec 2013 01:02:53 +0000 (17:02 -0800)
commita740576a4abf933de8f50787f24f24456cebd761
treee46ad30a727de5f0b8d91b381f1c67f05ebfb260
parentc5fe5d80680e2949ffe102180f5fc6cefc0d145f
x86: Slightly tweak the access_ok() C variant for better code

gcc can under very specific circumstances realize that the code
sequence:

foo += bar;
if (foo < bar) ...

... is equivalent to a carry out from the addition.  Tweak the
implementation of access_ok() (specifically __chk_range_not_ok()) to
make it more likely that gcc will make that connection.  It isn't
fool-proof (sometimes gcc seems to think it can make better code with
lea, and ends up with a second comparison), still, but it seems to be
able to connect the two more frequently this way.

Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/CA%2B55aFzPBdbfKovMT8Edr4SmE2_=%2BOKJFac9XW2awegogTkVTA@mail.gmail.com
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
arch/x86/include/asm/uaccess.h