From: Jaegeuk Kim <jaegeuk@kernel.org>
To: Hou Pengyang <houpengyang@huawei.com>
Cc: chao@kernel.org, linux-f2fs-devel@lists.sourceforge.net
Subject: Re: [PATCH 3/4] f2fs: check IPU before cp to avoid IPU being blocked by cp
Date: Tue, 25 Apr 2017 18:27:52 -0700 [thread overview]
Message-ID: <20170426012752.GA13449@jaegeuk.local> (raw)
In-Reply-To: <20170426012258.GA13347@jaegeuk.local>
On 04/25, Jaegeuk Kim wrote:
> Hi Pengyang,
>
> This makes xfstests/generic/013 stuck on fsstress.
Oh, it seems lock inversion problem between f2fs_lock_op and get_dnode_of_data.
The basic rule is get_dnode_of_data is covered by f2fs_lock_op.
I'll drop this patch at this moment.
Thanks,
>
> Thanks,
>
> On 04/25, Hou Pengyang wrote:
> > IPU checking is under f2fs_lock_op, as a result, some IPU page(such as fsync/fdatasync IPU)
> > may be blocked by a long time cp.
> >
> > This patch fix this by doing IPU checking before f2fs_lock_op, so fsync IPU could go along with cp.
> >
> > Suggested-by: He Yunlei <heyunlei@huawei.com>
> > Signed-off-by: Hou Pengyang <houpengyang@huawei.com>
> > ---
> > fs/f2fs/data.c | 14 +++++++-------
> > fs/f2fs/gc.c | 1 +
> > fs/f2fs/segment.c | 1 +
> > 3 files changed, 9 insertions(+), 7 deletions(-)
> >
> > diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
> > index 183a426..9bf9c7d 100644
> > --- a/fs/f2fs/data.c
> > +++ b/fs/f2fs/data.c
> > @@ -1380,19 +1380,20 @@ int do_write_data_page(struct f2fs_io_info *fio)
> > * it had better in-place writes for updated data.
> > */
> > if (need_inplace_update(fio)) {
> > - f2fs_bug_on(fio->sbi, !fio->cp_rwsem_locked);
> > - f2fs_unlock_op(fio->sbi);
> > - fio->cp_rwsem_locked = false;
> > -
> > + f2fs_bug_on(fio->sbi, fio->cp_rwsem_locked);
> > err = rewrite_data_page(fio);
> > trace_f2fs_do_write_data_page(fio->page, IPU);
> > set_inode_flag(inode, FI_UPDATE_WRITE);
> > } else {
> > + if (!fio->cp_rwsem_locked)
> > + f2fs_lock_op(fio->sbi);
> > write_data_page(&dn, fio);
> > trace_f2fs_do_write_data_page(page, OPU);
> > set_inode_flag(inode, FI_APPEND_WRITE);
> > if (page->index == 0)
> > set_inode_flag(inode, FI_FIRST_BLOCK_WRITTEN);
> > + if (!fio->cp_rwsem_locked)
> > + f2fs_unlock_op(fio->sbi);
> > }
> > out_writepage:
> > f2fs_put_dnode(&dn);
> > @@ -1473,13 +1474,12 @@ static int __write_data_page(struct page *page, bool *submitted,
> > if (!err)
> > goto out;
> > }
> > - f2fs_lock_op(sbi);
> > +
> > + fio.cp_rwsem_locked = false;
> > if (err == -EAGAIN)
> > err = do_write_data_page(&fio);
> > if (F2FS_I(inode)->last_disk_size < psize)
> > F2FS_I(inode)->last_disk_size = psize;
> > - if (fio.cp_rwsem_locked)
> > - f2fs_unlock_op(sbi);
> > done:
> > if (err && err != -ENOENT)
> > goto redirty_out;
> > diff --git a/fs/f2fs/gc.c b/fs/f2fs/gc.c
> > index e034857..b32cc30 100644
> > --- a/fs/f2fs/gc.c
> > +++ b/fs/f2fs/gc.c
> > @@ -715,6 +715,7 @@ static void move_data_page(struct inode *inode, block_t bidx, int gc_type,
> > .old_blkaddr = NULL_ADDR,
> > .page = page,
> > .encrypted_page = NULL,
> > + .cp_rwsem_locked = true,
> > };
> > bool is_dirty = PageDirty(page);
> > int err;
> > diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c
> > index 9f86b98..463a77b 100644
> > --- a/fs/f2fs/segment.c
> > +++ b/fs/f2fs/segment.c
> > @@ -293,6 +293,7 @@ static int __commit_inmem_pages(struct inode *inode,
> > .op_flags = REQ_SYNC | REQ_PRIO,
> > .old_blkaddr = NULL_ADDR,
> > .encrypted_page = NULL,
> > + .cp_rwsem_locked = true,
> > };
> > pgoff_t last_idx = ULONG_MAX;
> > int err = 0;
> > --
> > 2.10.1
> >
> >
> > ------------------------------------------------------------------------------
> > Check out the vibrant tech community on one of the world's most
> > engaging tech sites, Slashdot.org! http://sdm.link/slashdot
> > _______________________________________________
> > Linux-f2fs-devel mailing list
> > Linux-f2fs-devel@lists.sourceforge.net
> > https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
>
> ------------------------------------------------------------------------------
> Check out the vibrant tech community on one of the world's most
> engaging tech sites, Slashdot.org! http://sdm.link/slashdot
> _______________________________________________
> Linux-f2fs-devel mailing list
> Linux-f2fs-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
next prev parent reply other threads:[~2017-04-26 1:28 UTC|newest]
Thread overview: 10+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-04-25 12:45 [PATCH 0/4] optimize f2fs IPU v3 Hou Pengyang
2017-04-25 12:45 ` [PATCH 1/4] f2fs: reconstruct code to write a data page Hou Pengyang
2017-04-25 12:45 ` [PATCH 2/4] f2fs: lookup extent cache first under IPU scenario Hou Pengyang
2017-04-25 21:35 ` Jaegeuk Kim
2017-04-25 12:45 ` [PATCH 3/4] f2fs: check IPU before cp to avoid IPU being blocked by cp Hou Pengyang
2017-04-26 1:22 ` Jaegeuk Kim
2017-04-26 1:27 ` Jaegeuk Kim [this message]
2017-04-26 2:23 ` Chao Yu
2017-04-25 12:45 ` [PATCH 4/4] f2fs: unlock dnode before IPU Hou Pengyang
2017-04-25 21:27 ` Jaegeuk Kim
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20170426012752.GA13449@jaegeuk.local \
--to=jaegeuk@kernel.org \
--cc=chao@kernel.org \
--cc=houpengyang@huawei.com \
--cc=linux-f2fs-devel@lists.sourceforge.net \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).