From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1756755Ab1BJSIU (ORCPT ); Thu, 10 Feb 2011 13:08:20 -0500 Received: from mx1.redhat.com ([209.132.183.28]:60017 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1756654Ab1BJSIT (ORCPT ); Thu, 10 Feb 2011 13:08:19 -0500 Date: Thu, 10 Feb 2011 19:08:02 +0100 From: Andrea Arcangeli To: Dave Hansen Cc: linux-kernel@vger.kernel.org, linux-mm@kvack.org, Michael J Wolf , Johannes Weiner , David Rientjes Subject: Re: [PATCH 4/5] teach smaps_pte_range() about THP pmds Message-ID: <20110210180801.GA3347@random.random> References: <20110209195406.B9F23C9F@kernel> <20110209195411.816D55A7@kernel> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20110209195411.816D55A7@kernel> Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Wed, Feb 09, 2011 at 11:54:11AM -0800, Dave Hansen wrote: > @@ -385,8 +387,16 @@ static int smaps_pte_range(pmd_t *pmd, u > pte_t *pte; > spinlock_t *ptl; > > - split_huge_page_pmd(walk->mm, pmd); > - > + if (pmd_trans_huge(*pmd)) { > + if (pmd_trans_splitting(*pmd)) { > + spin_unlock(&walk->mm->page_table_lock); > + wait_split_huge_page(vma->anon_vma, pmd); > + spin_lock(&walk->mm->page_table_lock); the locking looks wrong, who is taking the &walk->mm->page_table_lock, and isn't this going to deadlock on the pte_offset_map_lock for NR_CPUS < 4, and where is it released? This spin_lock don't seem necessary to me. The right locking would be: spin_lock(&walk->mm->page_table_lock); if (pmd_trans_huge(*pmd)) { if (pmd_trans_splitting(*pmd)) { spin_unlock(&walk->mm->page_table_lock); wait_split_huge_page(vma->anon_vma, pmd); } else { smaps_pte_entry(*(pte_t *)pmd, addr, HPAGE_SIZE, walk); spin_unlock(&walk->mm->page_table_lock); return 0; } I think it worked because you never run into a pmd_trans_splitting pmd yet, and you were running smaps_pte_entry lockless which could race against split_huge_page (but it normally doesn't). > + } else { > + smaps_pte_entry(*(pte_t *)pmd, addr, HPAGE_SIZE, walk); > + return 0; > + } > + } > pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl); > for (; addr != end; pte++, addr += PAGE_SIZE) > smaps_pte_entry(*pte, addr, PAGE_SIZE, walk); > _ > From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail190.messagelabs.com (mail190.messagelabs.com [216.82.249.51]) by kanga.kvack.org (Postfix) with SMTP id BA9528D0039 for ; Thu, 10 Feb 2011 13:08:10 -0500 (EST) Date: Thu, 10 Feb 2011 19:08:02 +0100 From: Andrea Arcangeli Subject: Re: [PATCH 4/5] teach smaps_pte_range() about THP pmds Message-ID: <20110210180801.GA3347@random.random> References: <20110209195406.B9F23C9F@kernel> <20110209195411.816D55A7@kernel> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20110209195411.816D55A7@kernel> Sender: owner-linux-mm@kvack.org List-ID: To: Dave Hansen Cc: linux-kernel@vger.kernel.org, linux-mm@kvack.org, Michael J Wolf , Johannes Weiner , David Rientjes On Wed, Feb 09, 2011 at 11:54:11AM -0800, Dave Hansen wrote: > @@ -385,8 +387,16 @@ static int smaps_pte_range(pmd_t *pmd, u > pte_t *pte; > spinlock_t *ptl; > > - split_huge_page_pmd(walk->mm, pmd); > - > + if (pmd_trans_huge(*pmd)) { > + if (pmd_trans_splitting(*pmd)) { > + spin_unlock(&walk->mm->page_table_lock); > + wait_split_huge_page(vma->anon_vma, pmd); > + spin_lock(&walk->mm->page_table_lock); the locking looks wrong, who is taking the &walk->mm->page_table_lock, and isn't this going to deadlock on the pte_offset_map_lock for NR_CPUS < 4, and where is it released? This spin_lock don't seem necessary to me. The right locking would be: spin_lock(&walk->mm->page_table_lock); if (pmd_trans_huge(*pmd)) { if (pmd_trans_splitting(*pmd)) { spin_unlock(&walk->mm->page_table_lock); wait_split_huge_page(vma->anon_vma, pmd); } else { smaps_pte_entry(*(pte_t *)pmd, addr, HPAGE_SIZE, walk); spin_unlock(&walk->mm->page_table_lock); return 0; } I think it worked because you never run into a pmd_trans_splitting pmd yet, and you were running smaps_pte_entry lockless which could race against split_huge_page (but it normally doesn't). > + } else { > + smaps_pte_entry(*(pte_t *)pmd, addr, HPAGE_SIZE, walk); > + return 0; > + } > + } > pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl); > for (; addr != end; pte++, addr += PAGE_SIZE) > smaps_pte_entry(*pte, addr, PAGE_SIZE, walk); > _ > -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/ Don't email: email@kvack.org