From mboxrd@z Thu Jan 1 00:00:00 1970 From: Eric Sandeen Subject: Re: [PATCH] ext4: Resolve the hang of direct i/o read in handling EXT4_IO_END_UNWRITTEN. Date: Tue, 16 Aug 2011 09:06:12 -0500 Message-ID: <4E4A7954.8030603@redhat.com> References: <1313476673-4866-1-git-send-email-tm@tao.ma> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Cc: linux-ext4@vger.kernel.org, mjt@tls.msk.ru, tytso@mit.edu To: Tao Ma Return-path: Received: from mx1.redhat.com ([209.132.183.28]:43094 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752155Ab1HPOGT (ORCPT ); Tue, 16 Aug 2011 10:06:19 -0400 In-Reply-To: <1313476673-4866-1-git-send-email-tm@tao.ma> Sender: linux-ext4-owner@vger.kernel.org List-ID: On 8/16/11 1:37 AM, Tao Ma wrote: > From: Tao Ma > > EXT4_IO_END_UNWRITTEN flag set and the increase of i_aiodio_unwritten should > be done simultaneously since ext4_end_io_nolock always clear the flag and > decrease the counter in the same time. > > We don't increase i_aiodio_unwritten when setting EXT4_IO_END_UNWRITTEN so > it will go nagative and causes some process to wait forever. > > Part of the patch came from Eric in his e-mail, but it doesn't fix the problem > met by Michael actually. > http://marc.info/?l=linux-ext4&m=131316851417460&w=2 > > Reported-and-Tested-by: Michael Tokarev > Signed-off-by: Eric Sandeen > Signed-off-by: Tao Ma Yes, this looks right to me. It's the patch I meant to send, but I missed one spot in my haste. ;) Thanks. -Eric > --- > Eric talked about creating a helper for this and I feel that it looks > like a new functionality not a bug fix and decided to leave it to the > next merge window. > > fs/ext4/inode.c | 9 ++++++++- > fs/ext4/page-io.c | 6 ++++-- > 2 files changed, 12 insertions(+), 3 deletions(-) > > diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c > index d47264c..40c0b10 100644 > --- a/fs/ext4/inode.c > +++ b/fs/ext4/inode.c > @@ -2668,8 +2668,15 @@ static void ext4_end_io_buffer_write(struct buffer_head *bh, int uptodate) > goto out; > } > > - io_end->flag = EXT4_IO_END_UNWRITTEN; > + /* > + * It may be over-defensive here to check EXT4_IO_END_UNWRITTEN now, > + * but being more careful is always safe for the future change. > + */ > inode = io_end->inode; > + if (!(io_end->flag & EXT4_IO_END_UNWRITTEN)) { > + io_end->flag |= EXT4_IO_END_UNWRITTEN; > + atomic_inc(&EXT4_I(inode)->i_aiodio_unwritten); > + } > > /* Add the io_end to per-inode completed io list*/ > spin_lock_irqsave(&EXT4_I(inode)->i_completed_io_lock, flags); > diff --git a/fs/ext4/page-io.c b/fs/ext4/page-io.c > index 430c401..78839af 100644 > --- a/fs/ext4/page-io.c > +++ b/fs/ext4/page-io.c > @@ -334,8 +334,10 @@ submit_and_retry: > if ((io_end->num_io_pages >= MAX_IO_PAGES) && > (io_end->pages[io_end->num_io_pages-1] != io_page)) > goto submit_and_retry; > - if (buffer_uninit(bh)) > - io->io_end->flag |= EXT4_IO_END_UNWRITTEN; > + if (buffer_uninit(bh) && !(io_end->flag & EXT4_IO_END_UNWRITTEN)) { > + io_end->flag |= EXT4_IO_END_UNWRITTEN; > + atomic_inc(&EXT4_I(inode)->i_aiodio_unwritten); > + } > io->io_end->size += bh->b_size; > io->io_next_block++; > ret = bio_add_page(io->io_bio, bh->b_page, bh->b_size, bh_offset(bh));