From: Zhang Zhen Subject: Re: [PATCH] fs/ext4: fix potential endless loop in ext4_mb_discard_group_preallocations() Date: Thu, 17 Sep 2015 09:27:44 +0800 Message-ID: <55FA1710.7080602@huawei.com> References: <1441013790-87948-1-git-send-email-zhenzhang.zhang@huawei.com> <55E42270.9040503@huawei.com> <20150916194510.GC29530@quack.suse.cz> Mime-Version: 1.0 Content-Type: text/plain; charset="ISO-8859-1" Content-Transfer-Encoding: 7bit Cc: "tytso@mit.edu" , , "linux-ext4@vger.kernel.org" To: Jan Kara Return-path: Received: from szxga02-in.huawei.com ([119.145.14.65]:48882 "EHLO szxga02-in.huawei.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752825AbbIQB3H (ORCPT ); Wed, 16 Sep 2015 21:29:07 -0400 In-Reply-To: <20150916194510.GC29530@quack.suse.cz> Sender: linux-ext4-owner@vger.kernel.org List-ID: On 2015/9/17 3:45, Jan Kara wrote: > On Mon 31-08-15 17:46:24, Zhang Zhen wrote: >> In ext4_mb_discard_group_preallocations(), if free is always less than needed, >> and some PAs are always used in every loop, it will be endless loop. >> >> Here we pick a random value to limit the max number of loop. > > Were you able to trigger this in practice or is it just a theoretical > concern? > We found the problem in a stress test. It lead to system hungtask. Thanks! > My slight concern is that in theory we could prematurely declare ENOSPC > with this patch since ext4_mb_discard_preallocations() doesn't reliably > discard all the preallocations anymore. But probably that's acceptable. > But we should add a comment before ext4_mb_discard_group_preallocations() > saying that the functions needn't free all the preallocations. > > Honza >> >> Signed-off-by: Zhang Zhen >> --- >> fs/ext4/mballoc.c | 7 +++++-- >> 1 file changed, 5 insertions(+), 2 deletions(-) >> >> diff --git a/fs/ext4/mballoc.c b/fs/ext4/mballoc.c >> index 34b610e..553fbde 100644 >> --- a/fs/ext4/mballoc.c >> +++ b/fs/ext4/mballoc.c >> @@ -3836,6 +3836,7 @@ ext4_mb_discard_group_preallocations(struct super_block *sb, >> int err; >> int busy = 0; >> int free = 0; >> + int tried = 0; >> >> mb_debug(1, "discard preallocation for group %u\n", group); >> >> @@ -3886,9 +3887,11 @@ repeat: >> list_add(&pa->u.pa_tmp_list, &list); >> } >> >> - /* if we still need more blocks and some PAs were used, try again */ >> - if (free < needed && busy) { >> + /* if we still need more blocks and some PAs were used, try again, >> + here 20 is a ramdon value. */ >> + if (free < needed && busy && tried < 20) { >> busy = 0; >> + tried++; >> ext4_unlock_group(sb, group); >> cond_resched(); >> goto repeat; >> -- >> 1.9.1 >> >> >> . >> >> >> >>