From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1759678Ab0JVTI6 (ORCPT ); Fri, 22 Oct 2010 15:08:58 -0400 Received: from kroah.org ([198.145.64.141]:34231 "EHLO coco.kroah.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1759405Ab0JVS4j (ORCPT ); Fri, 22 Oct 2010 14:56:39 -0400 X-Mailbox-Line: From gregkh@clark.site Fri Oct 22 11:52:32 2010 Message-Id: <20101022185232.670425351@clark.site> User-Agent: quilt/0.48-11.2 Date: Fri, 22 Oct 2010 11:51:33 -0700 From: Greg KH To: linux-kernel@vger.kernel.org, stable@kernel.org Cc: stable-review@kernel.org, torvalds@linux-foundation.org, akpm@linux-foundation.org, alan@lxorguk.ukuu.org.uk Subject: [059/103] De-pessimize rds_page_copy_user In-Reply-To: <20101022185455.GA9114@kroah.com> Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org 2.6.35-stable review patch. If anyone has any objections, please let us know. ------------------ From: Linus Torvalds commit 799c10559d60f159ab2232203f222f18fa3c4a5f upstream. Don't try to "optimize" rds_page_copy_user() by using kmap_atomic() and the unsafe atomic user mode accessor functions. It's actually slower than the straightforward code on any reasonable modern CPU. Back when the code was written (although probably not by the time it was actually merged, though), 32-bit x86 may have been the dominant architecture. And there kmap_atomic() can be a lot faster than kmap() (unless you have very good locality, in which case the virtual address caching by kmap() can overcome all the downsides). But these days, x86-64 may not be more populous, but it's getting there (and if you care about performance, it's definitely already there - you'd have upgraded your CPU's already in the last few years). And on x86-64, the non-kmap_atomic() version is faster, simply because the code is simpler and doesn't have the "re-try page fault" case. People with old hardware are not likely to care about RDS anyway, and the optimization for the 32-bit case is simply buggy, since it doesn't verify the user addresses properly. Reported-by: Dan Rosenberg Acked-by: Andrew Morton Signed-off-by: Linus Torvalds Signed-off-by: Greg Kroah-Hartman --- net/rds/page.c | 27 +++++++-------------------- 1 file changed, 7 insertions(+), 20 deletions(-) --- a/net/rds/page.c +++ b/net/rds/page.c @@ -57,30 +57,17 @@ int rds_page_copy_user(struct page *page unsigned long ret; void *addr; - if (to_user) + addr = kmap(page); + if (to_user) { rds_stats_add(s_copy_to_user, bytes); - else + ret = copy_to_user(ptr, addr + offset, bytes); + } else { rds_stats_add(s_copy_from_user, bytes); - - addr = kmap_atomic(page, KM_USER0); - if (to_user) - ret = __copy_to_user_inatomic(ptr, addr + offset, bytes); - else - ret = __copy_from_user_inatomic(addr + offset, ptr, bytes); - kunmap_atomic(addr, KM_USER0); - - if (ret) { - addr = kmap(page); - if (to_user) - ret = copy_to_user(ptr, addr + offset, bytes); - else - ret = copy_from_user(addr + offset, ptr, bytes); - kunmap(page); - if (ret) - return -EFAULT; + ret = copy_from_user(addr + offset, ptr, bytes); } + kunmap(page); - return 0; + return ret ? -EFAULT : 0; } EXPORT_SYMBOL_GPL(rds_page_copy_user);