Return-Path: <linux-kernel-owner@vger.kernel.org>
Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand
	id S932219AbcCCDDs (ORCPT <rfc822;w@1wt.eu>);
	Wed, 2 Mar 2016 22:03:48 -0500
Received: from mailapp01.imgtec.com ([195.59.15.196]:39749 "EHLO
	mailapp01.imgtec.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org
	with ESMTP id S932191AbcCCDDq (ORCPT
	<rfc822;linux-kernel@vger.kernel.org>);
	Wed, 2 Mar 2016 22:03:46 -0500
Message-ID: <56D7A987.5040602@imgtec.com>
Date: Wed, 2 Mar 2016 19:03:35 -0800
From: Leonid Yegoshin <Leonid.Yegoshin@imgtec.com>
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:31.0) Gecko/20100101 Thunderbird/31.2.0
MIME-Version: 1.0
To: Paul Burton <paul.burton@imgtec.com>, <linux-mips@linux-mips.org>
CC: Lars Persson <lars.persson@axis.com>,
        "stable # v4 . 1+" <stable@vger.kernel.org>,
        "Steven J. Hill" <Steven.Hill@imgtec.com>,
        "David Daney" <david.daney@cavium.com>,
        Huacai Chen <chenhc@lemote.com>,
        "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>,
        <linux-kernel@vger.kernel.org>,
        Andrew Morton <akpm@linux-foundation.org>,
        Jerome Marchand <jmarchan@redhat.com>,
        "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>,
        Ralf Baechle <ralf@linux-mips.org>
Subject: Re: [4/4] MIPS: Sync icache & dcache in set_pte_at
References: <1456799879-14711-5-git-send-email-paul.burton@imgtec.com>
In-Reply-To: <1456799879-14711-5-git-send-email-paul.burton@imgtec.com>
Content-Type: text/plain; charset="utf-8"; format=flowed
Content-Transfer-Encoding: 7bit
X-Originating-IP: [10.20.3.92]
Sender: linux-kernel-owner@vger.kernel.org
List-ID: <linux-kernel.vger.kernel.org>
X-Mailing-List: linux-kernel@vger.kernel.org
Content-Length: 1289
Lines: 40

Paul Burton wrote:

> It is, however, used in such a way by others & seems to me like the only
> correct way to implement the lazy cache flushing. The alternative would
> be to adjust all generic code to ensure flush_icache_page gets called
> before set_pte_at

... which is an exact case right now. Both calls of flush_icache_page() are:

1) do_swap_page()

         flush_icache_page(vma, page);
         if (pte_swp_soft_dirty(orig_pte))
                 pte = pte_mksoft_dirty(pte);
         set_pte_at(mm, address, page_table, pte);
...

2) do_set_pte()

         flush_icache_page(vma, page);
         entry = mk_pte(page, vma->vm_page_prot);
         if (write)
                 entry = maybe_mkwrite(pte_mkdirty(entry), vma);
         if (anon) {
                 inc_mm_counter_fast(vma->vm_mm, MM_ANONPAGES);
                 page_add_new_anon_rmap(page, vma, address);
         } else {
                 inc_mm_counter_fast(vma->vm_mm, MM_FILEPAGES);
                 page_add_file_rmap(page);
         }
         set_pte_at(vma->vm_mm, address, pte, entry);

         /* no need to invalidate: a not-present page won't be cached */
         update_mmu_cache(vma, address, pte);
....

You should only be sure that flush_icache_page() completes a lazy 
D-cache flush.

- Leonid.